Reading Comprehension of Materials Written with Select Oral Language Patterns: a Study at Grades 2 and 4

DCCUMENT RESUME

ED 036 405 24 RE 002 386

AUTHOR TATHAM, SUSAN MASLAND TITLE READ-NG COMPREHENSION CF MATERIALS WRITTEN WITH SELECT ORAL LANGUAGE PATTERNS: A STUDY AT GRADES 2 AND 4. INSTITUTION WISCONSIN UNIV., MADISON. RESEARCH AND DEVELOPMENT CENTER FOR COGNITIVE LEARNING. SPONS AGENCY OFFICE OF EDUCATION (DHEW), WASHINGTON, D.C. BUREAU OF RESEARCH.. REPORT MO TR-86 BUREAU NO BR-5-0216 PUB DATE JUL 69 CONTRACT OEC-5-10-154 MOTE 1422..

EDRS PRICE EDRS PRICE MF-$0..75 HC-$7.20 DESCRIPTORS GRADE 2, GRADE 4, *LANGUAGE PATTERNS, *LANGUAGE RESEARCH, LANGUAGE ROLE, *ORAL EXPRESSION, *READING COMPREHENSION, *READING RESEARCH

ABSTRACT TO DETERMINE WHETHER OR NOT STUDENTS IN GRADES 2 AND 4 COMPREHENDED MATERIALS WRITTEN WITH PATTERNS THAT APPEAR FREQUENTLY IN THEIR SPEECH BETTER THAN MATERIALS WRITTEN WITH PATTERNS THAT APPEAR INFREQUENTLY, TWO READING COMPREHENSION TESTS WERE DEVISED BY THE INVESTIGATOR* SUBJECTS WERE ALI SECCND- AND FOURTH-GRADE CLASSROOMS FROM TWO SIMILAR SCHOCLS: 163 GRADE-2 STUDENTS (81 GIRLS, 82 BOYS) AND 137 GRADE-4 STUDENTS (69 GIRLS, 68 BOYS).. TEST A USED FREQUENTLY OCCURRING PATTERNS FRCM THE ORAL LANGUAGE OF SECOND AND FOURTH GRADERS, AND TEST B USED INFREQUENTLY OCCURRING PATTERNS IN THE ORAL LANGUAGE OF STUDENTS IRCM THE SAME GRADES. PATTERNS WERE SELECTED FROM STRICKLAND'S STUDY (1962). CHI SQUARE AND T TESTS WERE USED TO ANALYZE THE DATA RESULTS INDICATED (1) SIGNIFICANTLY MORE GRADE-2 AND -4 STUDENTS OBTAINED HIGHER SCORES ON TEST A THAN ON TEST B (P .091);(2) GRADE-4 STUDENTS PERFORMED SIGNIFICANTLY BETTER THAN GRADE-2 STUDENTS ON BOTH TESTS (P .01); (3) IN GENERAL, THERE WERE NO SIGNIFICANT SEX -DIFFERENCES ON EITHER TEST WITHIN OR ACROSS GRADES. TABLES, CHARTS, AND A BIBLIOGRAPHY ARE INCLUDED.(AUTHOR/WB) 4.1. S. DEPARTMENT OF HEALTH, OFFICE OF EDUCATION EDUCATION & WELFARE 11E002 384 TXISSTATEDPERSON DOCUMENTPOSITION DO OR NOT ORGMIZATI:N,OR HAS POLICY.NECESGAIZILY BEEN oRin10^71q" REPRODUCER EXACIINRE:'1:E.'',ENT RS 0; t!":71 OFFICE OF EDUCA I ION RECF,IVEDV!"1:7 FROM ON THE OPIN'fIN5 EDO 36405 /3-Y-0.2-/z72-e16 fg- 2- o 6-/X

Technical Report No. 86 READING COMPREHENSION OF MATERIALS WRITTEN WITH SELECT ORAL LANGUAGE PATTERNS: A STUDY AT GRADES 2 AND 4

Report from the Individually Guided Instruction in Elementary Reading Project

By Susan Masland Tatham

Wayne Otto,Professor of Curriculum and Instruction Chairman of the Examining Committee and Principal Investigator

Wisconsin Research and Development Center for Cognitive Learning The University of Wisconsin Madison, Wisconsin July 1969

This Technical Report is a doctoral dissertation reporting research supported by the Wisconsin Research and Development Center for Cognitive Learning. Since it has been approved bya Uni- versity Examining Committee, it has not been reviewed by the Center.It is published by the Center as a record of some of the Center's activities and asa service to the student. The bound original is in The University of Wisconsin Memorial Library. Published by the Wisconsin Research and Development Center for Cognitive Learning, supported in part as a research and development center by funds from the United States Office of Education, Department of Health, Education, and Welfare. The opinions expressed herein do not necessarily reflect the position or policy of the Office of Education andno official endorsement by the Office of Education should be inferred. Center No. C-03 / Contract OE 5-10-154 NATIONAL EVALUATION COMMITTEE Samuel Brownell Henry Chauncey Elizabeth Koontz Patrick Suppes Professor of Urban Education President President Professor Graduate School Educational Testing Service Yale University National Education Association Department of Mathematics Stanford University Launor F. Carter Martin Deutsch Senior Vice President on Roderick McPhee *Benton J. Underwood Director, Institute for President Professor Technology and Development Developmental Studies System Development Corporation Punahou School, Honolulu Department of Psychology New York Medical College Northwestern University

Francis S. Chase Jack Ed ling G. Wesley Sowards Professor Director, Teaching Research Department of Education Director, Elementary Education Division Florida State University University of Chicago Oregon State System of Higher Education

UNIVERSITY POLICY REVIEWBOARD Leonard Berkowitz John Guy Fowlkes Herbert J. Klausmeier Chairman Director M. Crawford Young Department of Psychology Director, R & D Center Associate Dean Wisconsin Improvement Program Professor of Educational The Graduate School Psychology

Archie A. Buchmiller Robert E. Grinder Donald J. McCarty Deputy State Superintendent Chairman Department of Public Instruction Dean Department of Educational School of Education Psychology *James W. Clr.ary H. Clifton Hutchins Ira Sharkansky Vice Chancellor k Academic Chairman Affairs Associate Professor of Political Department of Curriculum and Science Instruction

Leon D. Epstein Clauston Jenkins Henry C. Wein lick Dean Assistant Director College of Letters and Science Executive Secretary Coordinating Committee for Wisconsin Education Association Higher Education

EXECUTIVE COMMITTEE Edgar F. Borgatta Russell J. Hosier Wayne Otto Richard L. Venezky Brittingham Professor of Professor of Curriculum and Sociology Professor of Curriculum and Assistant Professor of English Instruction and of Business Instruction (Reading) and of Computer Sciences Max R. Goodson *Herbert J. Klausmeier Robert G. Petzold Professor of Educational Policy Director, R & 0 Center Associate Dean of the School Studies Professor of Educational of Education Psychology Professor of Curriculum and Instruction and of Music

FACULTY OF PRINCIPAL INVESTIGATORS

Ronald R. Allen Gary A. Davis Max R. Goodson Associate Professor of Speech Richard G. Morrow Associate Professor of Professor of Educational Policy Assistant and of Curriculum and Educational Psychology Professor of Instruction Studies Educational Administration

Vernon L. Allen M. Vere De Vault Warren 0. Hagstrom Wayne Otto Associate Professor of Psychology Professor of Curriculum and (On leave 1968.69) Professor of Sociology Professor of Curriculum and Instruction (Mathematics) Instruction (Reading)

Nathan S. Blount Frank H. Farley John G. Harvey Milton 0. Pella Associate Professor of English Assistant Professor of and of Curriculum and Associate Professor of Professor of Curriculum and Educational Psychology Mathematics and Curriculum Instruction (Science) Instruction and Instruction

Robert C. Calfee John Guy Fowlkes (Advisor) Herbert J. Klausmeier Associate Professor of Psychology Thomas A. Romberg Professor of Educational Director, R & D Center Assistant Professor of Administration Professor of Educational Mathematics and of Director of the Wisconsin Psychology Curriculum and Instruction Improvement Program

Robert E. Davidson Lester S. Golub Burton W. Kreitlow Richard L. Venezky Assistant Professor of Lecturer in Curriculum and Professor of Educational Policy Educational Assistant Professor of English Psychology Instruction and in English Studies and of Agricultural and of Computer Sciences and Extension Education MANAGEMENT COUNCIL *Herbert J. Klausmeier Thomas A. Romberg James E. Walter Dan G. Woolpert Director, R & D Center Director Director Director Acting Director, Program 1 Programs 2 and 3 Dissemination Section Operations and Business

Mary R. Quilling Director Technical Section * COMMITTEE CHAIRMAN STATEMENT OF FOCUS

The Wisconsin Research and Development Center for Cognitive Learn- ing focuses on contributing to a better understanding of cognitive learning by children and youth and to the improvement of related educational prac- tices. The strategy for research and development is comprehensive.It includes basic research to generate new knowledge about the conditions and processes of learning and about the processes of instruction , and the subsequent development of research-based instructional materials, many of which are designed for use by teachers and others for use by students. TheSe materials are tested and refined in school settings. Throughout these operations behavioral scientists, curriculum experts, academic scholars, and school people interact, insuring that the results of Center activities are based soundly on knowledge of subject matter and cognitive learning and that they are applied to the improvement of educational practice. This Technical Report is from the Individually Guided Instruction in Elementary Reading Project in Program 2. General objectives of the Pro- gram are to establish rationale and strategy for developing instructional systems, to identify sequences of concepts and cognitive skills,to identify or develop instructional materials associated with the concepts and cognitive skills, and to generate new knowledge about instructional procedures. Contributing to these Program objectives,the Reading Project staff, in cooperation with area teachers, prepared a scope and sequence statement of reading skills for the elementary school as a first step in the development of an instructional program. From this outline, assessment procedures and group placement tests have been developed, and existing instructional materials have been keyed to the outline. Research is conducted to refine the program and to generate new knowledge which will be incorporated into the system.

iii CONTENTS

List of Tables and Figures vii Abstract ix

IBackground of the Study and Related Literature 1 Background of the Study 1 Purpose of the Study 4 Definition .of Terms 5 Review of the Literature 6 Studies of Elementary School Children's Oral Language 7 Experimental Studies Using Written Material Based on an Analysis of Children's Oral Language 30 Summary 34 IIMethod 35 Overview of the Study 35 Development of the Reading Comprehension Tests 36 Selection of the Patterns 37 Determination of the Patterns to be Used 41 Control of Select Variables 44 Vocabulary Control 44 Content Control 45 Grammatical Complexity Control 47 Measurement of Reading Comprehension 51 Final Selection of Test Items 56 Composition of the Final. Tests 60 Subjects 61 Data Collection 64 Data Analysis 65 Summary 68 IIIResults and Discussion 69 Results of the Tests of Hypotheses 70 Results Related to Reading Comprehension of Select Oral Language Patterns 70 Each Grade as a Separate Group 72

iv(v Each Sex as a Separate Group Within Grades2 and 4 73 Results Related to Grade Level 74 Results Related to Sex 75 Summary of Results 84 Discussion of the Results 85 Results Related to Reading Comprehensionof Select Oral Language Patterns 85 Results Related to Grade Level 88 Results Related to Sex 91 Reliability of the Reading Comprehension Tests 92 Summary 96 Validity of the Reading Comprehension Tests 96 Construct Validity 97 Content Validity 101 Summary 102 Summary 103 IV Summary, Conclusions,and Implications 104 Background and Purpose 104 Related Research 105 Method 106 Results 109 Limitations 110 Conclusions 111 Implications 111 Implications for Beginning Reading Instruction 112 Implications for Teacher Training 112 Implications for Classroom Materials 114 Implications for Measurement of Language-Related Abilities 114 Implications for Research 115 Appendix A The Twenty-five Most Frequent Oral Language Patternsof Second and Fourth Graders as Identified by Strickland(1962)117 Appendix E Frequency Distributions of Scores on Test Aand Test B Within and Across Grades 2 and 4 124 Bibliography 137 Appendices B, C, D, and F are not printed herein. Seeoriginal manuscript in the University of Wisconsin MemorialLibrary.

vi LIST OF TABLES AND FIGURES

Table Page

1 The Rank Order and Frequency ofOral Language Patterns Selected for Test A 40

2 The Rank Order and Frequency of OralLanguage Patterns Selected for Test B 41

3 Index of Proportions of Frequent OralLanguage Patterns in Material Written for Second and FourthGraders 43

4 Index of Proportions of Infrequent OralLanguage Patterns in Material Written for Second andFourth Graders 43

5 "Hard" Words Used in Test A and Test B 46

6 Ratio of the Number of Different Words(Type) to the Total Number of Words (Token) in Test A and TestB 46

7 Analysis of Sentence Content in Tr'st A andTest B 48

8 Analysis of Picture Content in Test A andTest B 48

9 Total Index Ratings (IR) Given to Slots in theOral Language Patterns of Test A 51 10 Total Index Ratings (IR) Given to Slots in theOral Language Patterns of Test B 5 2

11 Summary of Total Index Ratings(IR) Given to Slots in Oral Language Patterns Common to Test A andTest B 53 12 The Mean, Range, and Standard Deviation of IQScores for All Subjects and for Subjects Grouped byGrade, Sex, and Sex within Grade 62 13 Proportion of Subjects in Three Levels Definedby the Duncan Socio-economic Index Scale 63

14 Chi Square Analyses of Reading ComprehensionTest Scores 71

vii Table Page 15 Mean, Range, Variance, and Standard Deviation for All 300 Subjects on Test A and Test B 7 2 16 Results of the t test of Mean Scores for Second and Fourth Graders 7 6 A Comparison of the Mean, Range, Variance, and Standard Deviation of Scores for Grades 2 and 4 77 18 Results of the t test of Mean Scores for Boys and GirlsAcross Grade Levels 78 19 Results of the t test of Mean Scores for Boys and Girlsin Second Grade 80 20 Results of the t test of Mean Scores for Boys and Girlsin Fourth Grade 8 2 21 Mean, Range, Variance, and Standard Deviation of Scoreson Test A and Test B for Girls and Boys Within and AcrossGrades Two and Four 83 22 Summary of the Results of the Tests of Hypotheses 86 23 The Number of Poor, Average, and Good ReadersAmong Fourth Grade. Girls and Boys as Determined by Scores fromthe Para- graph Meaning Section of the Stanford AchievementTest 88 24 Hoyt Reliability Coefficients Within and Across Gradesfor Test A and Test B 94 25 Partial Regression Coefficients of Test A andTest B 100 26 Partial Correlations of Test A and Test B withthe Paragraph Meaning Section of the Stanford Achievement Test 100 27 Validity Coefficient of Correlation betweenTest A, Test B, the Combined Scores, and the Paragraph MeaningSection of the Stanford Achievement Test 101

Figure

1 Sample Item from Test B 55 2 A Poor Distractor in the Pilot Study from TestB (the Middle Picture) 58 3 The Revised Distractor from Test B (the Middle Picture) 59

viii ABSTRACT

To determine whether or not second and fourth graders comprehend material written with oral language patterns that appear frequently in their speech better than material written with patterns that appear infrequently, two reading comprehension tests were devised by the investigator and given to all subjects.Subjects comprised all second and fourth graders in two similar schools; 163 second graders (81 girls,8 2 boys) and 137 fourth graders (69 girls, 68 boys) were used. Test A used patterns that appear frequently in second and fourth graders' oral language (e.g., noun- verb-object),and Test B used patterns that appear infrequently in the oral languag'e of the same grades (e.g., noun-verb-indirect object-direct object); patterns were selected from Strickland's study (1962).Chi square analysis was used to determine differential effects of frequent and infrequent oral language patterns on reading comprehension; t testswere used to determine differences between grade levels and sex. Results included the following:(1) significantly more second and fourth graders obtained higher scores on Test A than Test B (p < .001);(2) fourth graders performed significantly better than second graders on both tests (p < .01);(3) in general, there were no significant sex differenceson either test within or across grades.

ix I BACKGROUND OF THE STUDY AND RELATED LITERATURE

Background of the Study This study concerns the relationship between reading comprehension and material written with language patterns that appear in the oral language of second and fourth graders.Prior to 1959, conducting a study of this kind would have been a great problem, as there was no objective and reliable method for analyzing the speech of large numbers of children. Early attempts to study children's language were likely to be diaries of single children. Information collected in this manner was obviously not a basis for uroad generalizations. In 1959, however, a group of linguists and others devised a practical and useful method for analyzing children's speech. This scheme has been used in at least four major studies which provide information about the linguistic structure of children's oral language (Strickland, 1962; Hocker,

1963; Loban, 1963; Riling, 1965).In two of these studies,the newly acquired findings about children's oral language ability were related to children's, reading ability. This line of inquiry could very well have been prompted by the fact that elementary school teachers generally use a variety of methods to provide oral language experiences as a prerequisite to beginning reading instruction; yet very little is known about the nature of the assumed relationship between these two skills. The researchers found that the more mature users of the language, generally defined as those children who manipulated the basic patterns with greater flexibility, were also the better readers(Strickland, 1962; Loban, 1963).The findings suggest that facility with spoken language is related to successful comprehension of written language patterns. Since children's abilities to produce and comprehend spoken language are not evident to the same degree at a particular time(Vygotsky, 1962), it is likely that school children's ability to produce oral language patterns is,at some stage, also widely different from their ability td read these patferns. A reason for this discrepancy is provided by guists like Lefevre (1964), who have attempted to show that reading is not, contrary to popular definition, just "speech written down." Intona- tion qualities of speech that provide valuable meaning clues are absent in "speech written down"; the reader must learn conventions that are unique to written language such as capital letters and paragraphing. Whereas the process of learning to speak is apparently effortless and unconscious, the process of learning to read is obviously not effortless for many children. The enormous variety of teaching methods and materials and the concomitant controversy over their use attest to this statement. Several linguists and educators have discussed some of the unniN3- essary obstacles that the young reader must overcome before he learns how spoken language is represented in writing. Stevens (1965) criti- cized the use of "unnatural language" and "absurdly unnatural sentence rhythm" in most basal readers. Amsden (1964) suggested that children may fail in reading because of the use of teaching materials which are poorly designed in terms of their oral language patterns; he concluded that they may be reading a syntax that is too hard for them. Goodman (1968) has observed children inserting words as they read in an attempt to relate their customary way of speaking to the written syntax of basal

2 readers. Strickland (1962) found that there appeared to be no scheme for the development of control over sentence structures in the texts she analyzed; language patterns seemed to be introduced at random. Riling (1965) found that sixth grade textbooks do not use language in a way which is much superior structurally to the language of fourth grade textbooks. From observations like these, it is apparent that children's reading comprehension is often obstructed by materials that unnecessarily ob- scure the relationship between spoken and written languagea relationship which, according to linguists, is the crux of reading behavior (Carroll, 1964). Using materials structured more like the way children speak is a logical alternative to faulty comprehension caused by "unnatural language." Materials constructed this way would make use of an important principle from learr,ing theory, which defines efficient learning as that which "allows the learner to start with those learnings he brought with him to the task and upon these gradually build new responses which take him nearer and nearer to the desired terminal behavior [Carroll, 1964, pp. 348-91." In this case, the " learnings" a child brought with him would be his way of structuring what he has to say. The "new responses" would be his increasing ability to read language patterns that differ more and more from the ones he uses in his oral language. One study has focused on the possibility of improving children's reading comprehension by using materials structured more like theway they speak (Ruddell, 1963).Its major finding was that fourth graders comprehended material written with pattern'S that appear frequently in children's oral language significantly better than material written with patterns that appear infrequently in the oral language of children. More research is needed to confirm these results for different groups of children and at different grade levels.Determining whether or not certain

3 language patterns used by children are easy or difficult forthem to comprehend in written form could bring a relevant but heretoforeneglected characteristic of the readerhis oral languageto the conceptof readability. Research in this area could thus begin to define the natureof the relationship between oral language facility and reading ability.

Purpose of the Study The present study was undertaken to investigate further the relationship between reading comprehension and material written with select oral language patterns at second and fourth grades .The relation of sex to reading comprehension of these patterns was of additional interest, as it was hoped that this study couldcontribute to the dialogue between those who have found significant sex differences in language ahilities of children and those who have not. Comparing the performance of boys and girls at two grades was intended to encourage the emergence of any developmental levels with respect to children's ability to comprehend diverse kinds of language patterns. Fourth graders were included as subjects so that a comparison with the findings of the other study in this area would be possible. Second graders were added to provide some information about children who areconsiderably less experienced in relating spoken and written language. The study was designed to answer three basic questions: Ql: Do significantly more second and fourth graders comprehend material written with frequent oral language patterns better than material written with infrequent oral language patterns? Q 2: Do fourth graders comprehend material written with frequent and infrequent oral language patterns significantly better than second graders? Q3: Do girls comprehend material written with frequent and infrequent oral language patterns significantly better than boys?

4 Definition of Terms

Several terms are used frequently throughout the following chapters. Since definitions are often confined to the context of the study, theterms are listed here for the reader's convenience. Language pattern. --A language pattern isa unit in oral or written language that indicates the sequence of slots and movables.Examples are the noun-verb-object pattern, coded as 1 2 4 (She jumps the log), and the noun-verb-adverb-of-mannerpattern, coded as 1 2 M2 (She jumps quickly). Sentence.A sentence is a single languagepattern, or several patterns joined by a conjunction; that begins witha capital letter and ends with a period. Phonological unit. Used in the analysis of children'soral language, this unit is determined by the inflection,stress, and pause of the speaker; oral language is segmented into phonologicalunits according to the sound- system of the English language (Loban, 1963). Thisterm has been applied inaccurately asa synonym for "sentence." T-unit. Popularized by Hunt (1966) in hiswork with analyses of children's written language, this term isdefined as a main clause with any attached subordinate clauses. The T-unit is considereda more accurate and thus more meaningfulmeasure than the phonological unit. Slot. A slot is an immovable unit ina pattern.In this study, these units are designated in the followingmanner:1 = subject; 2 = verb; 2b = passive verb, verb of theto be class, or copulative verb; 3= indirect object; 4= direct object; 5 = predicate nominative-; Movable. A movable isa unit of a pattern that generally does not have a fixed position. The movablesin this study have been categorized into three subgroups:M1 = an adverb of place (e.g.,in the house;

5 here); M2 = an adverb of manner (e.g.,quickly; on foot); M3 = an adverb of time (e.g.,in winter; then). T marker. A T marker used before a pattern indicates that the pattern is preceded by another pattern. A T marker used after a pattern indicates that the pattern is followed by another pattern.Joining two patterns with T markers, 12 4 +T and T 12 4, would result in 1 2 4 + 12 4, illustrated by the sentence, "The dog sees the cat and the cat sees the dog." Frequent oral language aattem. For purposes of the study, this type of pattern appeared among the ten most frequent patterns in the oral language of second and fourth graders as identified by Strickland (1962). Infrequent oral language pattern. An infrequent pattern was one that ranked lower than tenth in fr6quency in the oral language of second and fourth graders as identified by Strickland (1962). Slot filler. A slot filler comprises the single words, adjectives, phrases, or clauses with which a slot may be filled. Stanford subtest. This term refers to the Paragraph Meaning section of the Stanford Achievement Test taken by the subjects.

Review of the Literature The literature reviewed in this chapter is divided into two parts. The first section, studies of elementary school children's oral language, is limited to investigations that have been based on a broad analysis of oral language obtained from relatively large samples of elementary school children. Studies of children's language development have been reviewed in the Encyclopedia of Educational Research (Carroll, 1960), the April issue of the Review of Educational Research (Carroll, 1958;

6 Kjeldergaard, 1961; and Carroll, 1964), and a more recent publication by O'Donnell, Griffin, and Norris (1967). The intent in this chapter is to focus on studies that have used a more sophisticated and reliable method of analyzing children's oral language than existed before 1959. The number of studies is consequently smallfive. The second section, experimental studies using written material based on an analysis of children's oral language, is confined to a discussion of a single study, which is believed to be the only one in this area at present.

Studies of Elementary School Children's Oral Language In this section, the five broad analyses of children's oral language reported since 1959 are reviewed. The first study of its kind was published by Ruth Strickland in 1962. She undertook to study oral language samples of children from first through sixth grades with these purposes: (1) to analyze the structure of the oral language, (2) to compare the structure of the oral language with that found in children's texts, and (3) to determine whether or not characteristics of the oral language structure were related to the quality of a child's reading ability. The sample comprised 575 children who were randomly selected from the 16 public schools in Bloomington, Indiana. There were 100 subjects in each of Grades 1-3, 5, and 6, and 75 subjects in the fourth grade. The variables of age, sex, intelligence, and socioeconomic status were studied at each grade level, and the variable of parental education was considered in most of the grades.Some additional analyses were made at the sixth-grade level; children's oral language was compared to their silent reading comprehension, their listening comprehension, and their oral reading interpretation. The methodology and findings for each of

7 the three major, relevant concerns of the study will bediscussed separately below. All of the oral language samples were obtained by havingchildren come to a testing room at the school in groupsof twos and threes. The pupils sat at a table which had several familiar storybookcharacters on it. The examiner encouraged them to talk freely about anythingthat interested them; the figures and the examiner served to stimulate the conversation. The conversations were taped until each child was "talking naturally." The language samples were then transcribed, and 25 sentences or phonological unitsl from each child's sample were analyzed for these characteristics:(1) syntactic structure of sentences,(2) frequency of occurrence of certain language patterns,(3) amount and kinds of subordination, (4) length of sentences, and (5) flow of language. The relationship of select language variables to age, sex, and intelligence levels of the children was then ascertained.

Strickland's study, the term sentence meant phonological unit, which she defined as a "unit of speech ending with a distinct falling intonation which signals a terminal point [p. 16)." The unreliability of this unit, however, has been demonstrated by researchers like Kean (1967) and Riling (19 65),who attempted to explain how she distinguished between long phonological units joined by and and a series of short phonological units also joined by and. Her decision was based on (1) the pitch which marked the beginning and end of the unit, (2) the length of the pause before and, (3) the location of the falling intonation, and (4) the pronunciation of and. Some of these criteria are surely difficult to judge. A more reliable unit, the T-unit, has been developed and tested by Hunt (1965) in his studies of children's structures in written language. This unit consists of "one main clause with all the subordinate clauses attached to it [p. 20]." The use of this measure should provide more reliable results for future studies of children's oral language,

8 A singular feature of this study at the time of publication wasthe scheme of analysis used on the 14,375 phonological units that comprised the-language sample. The scheme was the product of linguists who met for this purpose in 1959 at IndianaUniversity.2 Two distinct kinds of analyses were outlined: Level I, in which sentences are divided into fixed slots (the immovable units in a language pattern such as the subject and verb) and movables (the unit of a pattern that generally does not have a fixed Position), and Level II, in which sentences areanalyzed for the type of subordinations used in tke slots and movables.3 Since Strickland's study was primarily a description of the oral language patterns , the findings were presented intables which summarized the frequency with which certain patterns appeared for select groups of children. Tests of significance to determine trends across the grades were not made; it is usually up to the reader to draw his own conclusions about the meaning of figures in the tables. The only statistic applied to the data was chi square, which was used to study the relationship between certain language variables and background data of the subjects. In sum, Strickland's study was not a statistical analysis, but rather a descriptive one. The findings should be considered in this light. Some of the major findings,of her analysis of children's oral language are summarized briefly below: 1. Data for the word-length of phonological units were presented for Grades 1,4, and 6 in two tables.In one, the length of phonological units

2 The participants were John Carroll, W. Nelson Francis,David Reed, Fred Householder, Harold Whitehall, Walter Loban, Virginia Mini, Mansur Ekhtiar, Eldonna Everetts, and Ruth G. Strickland. 3For a full description of these levels, the reader is referred to pp. 18-22 of Strickland's study (1962).

9 was summarized in terms of meanand range by grade and sex; differences between means were not tested for significance.In the second table, the mean length of phonological units for the highest andlowest quarters .. of pupils in Grades 2 and 4 was summarized according to age,intelligence, mental age, occupational status,and father's and mother's education. As no tests of significance between means were made,the conclusions were stated in a general way:(1) Sentences varied considerably

.. in length, the range at first grade being from 1-76 words,for example. Differences between means and variables of the subjects' background at Grades 4 and 6 were small but always in favor of the high group. (3) Length varied more within a grade than from grade to grade. 2. With respect to the occurrence of language patterns, Strickland found that the number ranged from 658 in the sample for first grade to 1 ,041 for the sixth grade. Although a wide range of language patterns was used at all grade levels , someappeared much more frequently than others at all the grade levels. Strickland suggested that therelatively few' language patterns are the "basic building blocks" of children's language. A table of the rank order and frequency of the structural patterns most commonly used by children waspresented for Grades 1, 2,, 4, and 6. Of the 5 highest rankingpatterns that were common to each of these grades, none contained a movable element. 3. Results of the Level II analysis were presented in a tablewhich .. indicated, for Grades 1, 2, 4, and 6, the per cent of use of each typeof subordinate element and the per cent of the sample at each gradelevel that used the element.It is clear from the table that children at all grade levels expanded and elaborated their sentences throu4h, the use of movables and elements of subordination. Differencesbetwe6n grade levels were not tested for significance, or evendiscussed, so the presence of

10 developmental levels was not readily apparent if, indeed, these levels exist in the data. 4. By means of chi square tests, the relation between the use of movables and subordination and select variables was ascertained. Dif- ferences significant at the .05 level or better were found between the use of subordination and movables by subjects in Grades 1, 4, and 6 who differed in verbal intelligence, mental age, occupational status, and parents' education. In sum, these generalizations can be derived from Strickland's analysis of children's oral language: 1.Children at all grade levels use a wide range of language patterns. 2.Certain patterns which children use with great frequency appear to be basic to their language. 3.Children at all grade levels can expand and elaborate their sentences through the use of movables and elements of subordination. 4. There are significant differences between the use of movables and patterns of subordination and variables of intelligence, mental age, occupational status, and parental education. A second major focus of the study was the comparison of children's language with the language of their textbooks.For this purpose, Strick- land selected four sets of readers for analysis. Each preprimer and eight pages from the other books comprising each set were studied to find out which of the patterns that were used most frequently by children in Grades 1, 2, 4, and 6 appeared in the texts for those grades. Two points need to be made to place the findings in perspective.First, certain patterns were selected from children's language and compared with patterns in the texts. No attempt was made to report those patterns in the text that did not coin- cide with these selected oral language patterns. Second,as Strickland was careful to state, this analysis was not meant to be a frequency

11 tabulation of the patterns in thebooks. The intent was merely to state whether or not certain patterns hadappeared at all on the sample pages. The findings were presented in atable which shows the rankorder of the selected frequent orallanguage patterns for each gradelevel and whether or not the pattern appeared inthe texts analyzed for that grade. Within the limitations of the samplingprocedures,these conclusions were drawn: 1. The patterns which appeared inthe sample were not the same from book to book either within or across anyof the series. 2. The patterns appeared to beintroduced somewhat at random in the texts. 3. There appeared to be no schemefor the development of control over sentence structuresin the texts that parallels thegenerally accepted scheme for the development of control overvocabulary. A third major concern of thestudy was to determine the relationship between oral language and certain readingabilities.A study of the relationship between orallanguage and reading achievement was made for Grades 2 and 6 only. The ParagraphMeaning test from the Gates Primary Reading Tests (Type AWR, Form 3) wasused to provide information about the reading comprehension of the 100second graders. The scores obtained from this test were translated intoreading age. Subjects were grouped into one of five categories ofreading age, and a tabulation of the frequency and average use of eachof the most frequent language patterns was made. Other comparisonsbetween the groups of children were made in terms of sentencelength, the use of movables , and the use of subordination.The results were presented in severaltables, and a few casestudies of representative children wereadded. Pertinent findings included the following:

12 1.Children in the lowest category of reading age werecharacterized by greater use of the basic patterns thanchildren in other categories; the basic patterns contained no movables . 2. Children whose reading scoresplaced them above the grade norm tended to use longer sentences. 3. When the children in the lowest andhighest categories were compared, there were no significantdifferences in their use of movables or their use of subordination patterns according to thechi square analysis performed. At the sixth-grade level, three measuresof comprehension were obtained:(1) silent reading comprehension ascertained by the Paragraph Meaning test of the Stanford Achievement Test(Intermediate Level, Form L), (2) oral reading interpretation defined by scores from the oral partof Sec- tion IV, Form A, Grades 1 to 8, of the Diagnostic ReadingTests (This test was administered individually and theoral reading was taped.), and (3) listening comprehension measured by a group test in whichchildren lis- tened to taped selections and answered written questionsafterward. The sixth graders were grouped on the basis of their gradeplacement scores on the silent reading comprehensiontest. The frequency with which select language patterns appeared for each of the resultingfive groups was presented in atMale. Similar tables were presented for subjects grouped by oral reading interpretation, which was ratedalong a five- point scale devised by Strickland for the study, and bylistening comprehension, which was measured by the number of correct responses tothe listening test.Comparisons of mean sentence length, the use of movables, and the use of subordination were made for those subjects who wererated high or low on any of the three comprehension tasks .Results of the chi square analyses were presented for thelast two comparisons between groups .

13 The findings, can be summarized as follows: 1. There was a relationship between the structureof children's oral language and silent reading comprehension. Pupils whoranked high in silent reading4. made greater use of movables and elements ofsubordina- tion than children who ranked low. This difference between groups was significant at the .02 level.Pupils who ranked high also had a greater mean sentence length than children inthe low group, but this difference was not tested statistically. 2. There was a relationship between oral reading interpretationand the structure of children's oral language. A difference significant at the .01 level was found between the high and low groups in oralreading interpretation and the use of movables and subordinate elements. 3. There was a relationship between listening comprehension and the structure of children's oral language. Children who were rated high in listening comprehension had a greater average use of movables and subordinate elements and a longer mean sentence length than children ranked low, but these differences were not statistically significant. 4.Children ranked high in silent reading, oral reading, and listening used longer sentences than children ranked low in these areas .The differences,however, amounted to 2.31 words at most. Strickland's work has been criticised for its restricted child population, the attempt to compare informal conversation with edited reading material (Burrows,1964), and its failure to differentiate details of subordination at the Level II analyses performed (Mayer, 1964).In general, however, this study has been marked as a significant contribution to descriptions of elementary school children's oral language. Evidence of the effect of Strickland's work (1962) on subsequent research is clear.In the years immediately after its publication, two

14 studies emerged which distinctly followed her methodology. The first was a master's thesis by Mary Elsa Hocker (1963). Her study of children's oral language was undertaken with the assumption that "the more nearly the printed material resembles the child's oral language patterns, the easier it should be for him to anticipate meaning in the sentences he is beginning to read [p. 1)." Being interested in the beginning stages of reading, she focused on about 40 first graders from two schools at an Air Force Base in Arizona. Her methodology was unique in that language samples were collected in a variety of situations.Birthday parties, Sunday school, general play, and classroom show-and-tell provided some of the settings in which she workeda distinct break away from the more structured situation used by Strickland (1962) and later by Loban (1963) and Riling (1965).Consecutive utterances of the subjects were recorded by the examiner who used the stenographic method and tape recorder, the latter when possible. A total of 2500 phonological units were collected and analyzed for (1) language pattern, (2) frequency of occurrence of these patterns,(3) length of sentences, (4) the kind of vocabulary used, and (5) interests. The first three concerns were also important to Strickland's work, while the last two were added by Hocker. As in Strickland's study, a primary concern was the segmentation of phonological units into distinct language patterns.For this purpose, then, Level i of the linguistic scheme previously discussed was used.It will be remembered that this level provides for segmenting any phonological unit into its slots (the fixed elements of a sentence) and movables (the sentence elements that generally do not have a fixed position). The findings were presented in a series of tables, one of which listed each pattern and the frequency with which it occurred. Statistics were not applied to the data so that, as in the greater part of Strickland's work, the

15 results were broad descriptions of general trends rather than statistical

evidence of trends. .. Among those findings that are most pertinent to the present study are the following: 1. The mean sentence length of the language sample was 4.86 words, which contrasts with 6.6 for the six-year-olds of Templin's study (1957) and 11.04 and 10.70, respectively, for the first grade boys and girls of Strickland's study (1962). 2. Of the five basic sentence patterns that occurred in the sample, three of them were dominant and formed the syntactic structure for nu- merous variations in the children's oral language:12 4 (nounverb object),1 2b 5 (nounto be verbpredicate nominative) , and 12 (noun verb).Although the first two patterns were among the most frequent in Strickland's sample at all grade levels (1962), the 12 pattern was generally considerably less frequent. 3. The trend of the pattern frequency, moved from the simple to the more complex in all cases. This finding supports Strickland's more detailed observations; the most frequent patterns were those without movables. Though more humble than Strickland's work, Rocker's study contrib- utes several important points that have not been developed to any extent beyond theory by other researchers in the area.First is her conclusion that the kind of oral language a child uses varies with the situation. Hocker apparently made this statement after considering the variance of sentence length of her subjects from situation to situation and the contrast between a child's verbal encounter with an adult and with another child.Unfortunately, a thorough investigation of the noted differences was not made.It is logical to assume that slight, or even significant,

t differences in the findings of studies of children's oral languageneed to be considered with respect to (I) the kind of stimuli used inobtaining the language sample, and (2) the situation in which the oral language was sampled. A second conclusion grew from observations ofdifferences in the language depending on where the language was used. Hockersuggested that because the language patterns used are influenced by the situationin which the child is presently operating, it would "be a mistake" toprovide him with a description of a school situation in the language patternsof the playground. Hocker did not differentiate between playground language patterns and schoolroom patterns explicitly enough to supportthis gen- eralization. The idea, however, should be considered in future research in which the relationship between children's oral language andwritten material is studied. Thus,although Hocker used the same scheme as Strickland for analyzing the oral language, unlike Strickland she varied the context inwhich the language samples were collected, and she limited her investigation to first graders. Perhaps the major contribution of this work is the attempt to begin answering a question posed in Strickland's study:Would children be "aided or hindered by the use of sentences in their books more like the sentences they use in their speech [1962, p. 106]?" Riling's study (1965) followed Strickland's methodology closely. The influence of the earlier work was made explicit in Riling's statement of purpose: This study is designed to analyze the structure of children's written and spoken language in the fourth and sixth grades; to ascertain the influence of certain variables such asintel- ligence, sex, socioeconomic status of the family, education of the parents,and ethnic origin on the form of the language of these children; and to compare the results of this study

17 with those obtained in a similar study conducted by Ruth G. Strickland of Indiana University [p. 3]. Riling was also interested in comparing the patterns which children use at these grade levels with those found in reading textbooks designed for thesegrades. A total of 300 children were drawn from six communities within a radius of 70 miles of Southeast Oklahoma State College; the sample comprised 114 Caucasian and 110 Negro children for Grade 4 and 110 Caucasian children for Grade 6. The composition of the sample contrasts with Strickland's (1962) undifferentiated group of 75 subjects in Grade 4 and 100 in Grade 6.The procedure for obtaining the language samples differed from those of both Strick- land and Rocker (1963). Perhaps to avoid criticism in comparing oral and textbook language, Riling tried to keep the narrative style of the texts that were to be analyzed constant with the narrative oral and written language of the children. Oral language was stimulated by a picture that was placed before the child, who came into the testing room alone and told his story about the picture; the written language sample was stimulated by a different picture that was used by the investigator in a regular classroom setting. The children discussed the picture and then wrote a story about it. The process of taping and analyzing the language was similar to Strickland's method. Twenty-five "independent verbalizations of oral ...and written language" (the number also selected by Strickland) were the maximum amount of language analyzed for each child. For her comparison of children's language with the language of textbooks, Riling designated pages 1, 10, 20, and 50 from the front of each book and four pages in the same relative order from the back as the sample. She followed this procedure for six commonly used textbooks for each grade. The sentences were analyzed by the two-level scheme previously

18 discussedLevel I showing the arrangement of basic grammatical elements and Level II showing elaboration of theseelements. To obtain a measure of reading comprehension, theStanford Achievement Test (also used in Strickland's study) was given to the subjects. Among the conclusions that are relevant to the present study are the following:

1.Children with higher intelligence, more highly educated parents, and higher family socioeconomic status do not show any significant su- periority in the use of the' simplest, most basic language structures over children less endowed. 2. Fluency of the use of language does not seem to be a reliable index of maturity of the use of lariguage. 3. Children use great structural variation in their oral language which they are unable to carry over into their written language. 4. Even as early as fourth grade, in narration children use the chief structures of the English language proportionately in a way that approaches the use of the adults who write nonfiction for books , magazines , etc. 5. The written language of the children of this study is inferior structurally to that of most of their textbooks, but not much.It is much more structurally varied than that of the textbooks designedfor slow readers. 6. Sixth-grade textbooks do not use language in a way which is much superior structurally to the language of fourth-grade textbooks. 7. None of the textbooks give attention to consistent development of sentence structure. 8. There seems to be some relationship between the use of clauses and phrases and the paragraph comprehension of children in Grade 4; but this relationship is not clear at the sixth-grade level.

19 In general, Riling found that the oral language patterns used by the children in her sample were quite similar to those reported by Strickland (1962). Differences could be due to any number of these factors:(1) the composition of the samples was not the same; (2) the stimulant to conversation was not the sameStrickland sought open discussion from two or three children while Riling steered the oral response from children by having them.tell a story about a picture; (3) the setting was not the same other children were the focus in Strickland's tape sessions while just an adult was present at Riling's;(4) the language samples were collected in different parts of the country (Indiana and Oklahoma). Despite some expected differences among the studies,Riling's work strengthens some of the tentative findings of both Strickland (1962) and Hocker (1963). First, relatively large samples of children's oral language can be recorded and analyzed according to the two-level scheme devised in 1959 (Strickland, 1962; Hocker, 1963; and Riling, 1965). Second, children's textbooks do not give much attention to a consistent development of sentence structure (Strickland, 1962; Riling, 1965). Third, it seems as though the kinds of structures that children use are affected by certain aspects of the situation in which the oral language is produced (Hocker, 1953; Riling, 1965). As previously discussed, Hocker emphasized the relative degree of informality and formality about the location of the child's speechschoolroom vs.playground. Riling looked at a slightly different aspect by emphasizing the purpose for the child's speech; she suggested that further research should consider the association between structure and purpose of the language. In a work that differed considerably from the methodology of the three studies discussed above, Lobeln (1963) reported on the longitudinal analysis of language used by children in kindergarten and the six years of elementary school. This study was continued through high school,

20 but only the section relevant to elementaryschool is discussed here. Loban was particularly Concerned with(1) defining the relationship among children's use of oral, written, listening, and readinglanguage, (2) developing fundamental methods of analysis in the studyof children's language, (3) locating significantfeatures of language that are worthy of further study and (4) determining predictable stages of growthin language. For the study, eleven kindergarten classrooms werematched with family backgrounds that were similar to the general population inthe city of Oakland, California; stratification was based on sex,racial background, socioeconomic status, and intellectual ability. A total of 338 children were included in the sample .4 Each year, samples of speech, writing, and reading were collected for each child under controlled conditions that were similar for all subjects . In addition, more extensive information was gathered for two subgroups which represented extreme deviations from the mean: a group of 30 subjects who were very high in language ability, and a group of 24 subjects who were very low in language ability. The criterion for determining the subgroups was two or more standard deviations from the mean of two measures which were given equal weight, vocabularyand teacher ratings. Although Loban was concerned with children's written language, the following discussion is limited to those pILocedures and findings that pertain directly to knowledge of children's oral language and its relation to reading. The procedure for collecting the oral language samples was more like Riling's (1965) than Strickland's (1962) or Hocker's (1963) in that children

4Insimilar studies, Strickland (1962) used 575 children in Grades 1-6 and Riling (1965) used 300 in Grades 4 and 6.

21 were interviewed individually.The examiner played a more definite role in stimulating conversation than in Riling's study, however, as the interview was divided into two parts:(1) the children were asked about "playmates, games, television, illness, and wishes [p. 3]"; then (2) the children were. shown a set of six pictures (the same ones being used for all subjects) and asked to discuss "what they saw in each picture and what they thought about each picture [p. 4]." The responses were recorded on an Audograph and then transcribed into typewritten form for analysis. The oral language samples were analyzed according to the two-level scheme previously discussed, in which the speech sample was segmented by intonation patterns and then divided into syntactic units which were examined for types of subordination. Since Loban has stated that all communication units (what Strickland called "patterns of language structure") can be classified as one of nine patterns,he limited his tabulation of pattern-types to the following list:

Pattern Symbol Examples one 1 2 or 1 0 Mary eats.(or)Mary is home. two 1 2 4 Mary eats strawberries. three 1 ® 5 Strawberries are berries. Strawberries are good. four 12 3 4 Mary threw the dog some biscuits. five 1 2 4 6 They elected Mary president. They thought Susie conceited. six (1)(2) 1 Here is Mary. There are four houses on Lime Street. seven Questions How does he do it?Is he here? eight Passive forms Strawberries were eaten by Mary. nine Requests, commands Go home.(or)Let us go home. (ten) Partials Any incomplete unit.(This is not actually a pattern like the preceding nine patterns .)[pp. 14-15]

22 From inspection of the list, it is clear that the distinctions between 1 2 M2 and 12 made in Strickland's work (1962) were not made in his analyses; the former pattern was considered an elaboration of the latter. In addition to this gross analysis of the language samples, six analyses were performed:5 1.Classification of speech in terms of its function. Eight categories were used for this purpose: facts and unelaborated perceptions,interpretations,personal associations, tentative statements or suppositions , generalizations , irrelevancies , direct questions, and figurative language. As Loban pointed out, his scheme was useful for this particular study, but under other data-gathering situations, it might not be comprehensive enough. 2.Classification of oral language style. In the study, eleven features of oral language style were identified and considered along a continuum which was scaled from 1 to 3: fluent to halting deliberate to impulsive coherent to incoherent, disorganized energetic to listless, weak, tired laconic to loquacious expressive to flat, expressionless mature to babyish distinct, clearly articulated to blurred, indistinct, mumbled conventional in usage to unconventional in usage ready in response to slow in response relaxed to tense, strained [P. 17] Some inconsistencies in the scale can be noted. Most of the language features in the left column are clearly positive attributes which move along

5Analyses 1, 5, and 6 were made only on data from the two subgroups. Analysis 3 was made only on data from the two subgroups and a small random sample of the total group.

23 the continuum to clearly negative characteristics in theright column. Two of the terms in the left column, however, are notcategorically either positive or negative.Is "laconic" a more desirable trait than "loquacious"? Does being "conventional in usage" (rated a 1, the highest score) mean being better than "unconventional usage" (rated a 3, the lowest score) ? Perhaps the latter includes the use ofmetaphor or im- agery, which is not generally consideredconventional. Loban may have clarified these points to those who were responsible for rating the children's languaae. 3. Amount of subordination.

4.Difficulties with conventions in usage and grammar. These categories for nonstandard usage, syntax, and grammar were items involving (1) the use of verbs, pronouns , prepositions , and con- junctions, modifiers, or nouns, and (2) omission or repetition. 5. Vocabulary measured by word frequencies. 6. Vocabulary measured by diversity. In addition, several ratings were obtained. Each year, the teachers rated the subjects on these language measures:(1) amount of language, (2) quality of vocabulary, (3) skill in communication, (4) organization, purpose, and control of language,(5) wealth of ideas, and (6) quality of listening. 6 An index of reading ability was obtained by having the investigator assign weights to each book that the child completed in a

6These measures are clearly subjective in nature. Where , for example, was "amount of language" ratedin the classroom or on the playground? For some children, there may be a drastic reduction in langu:_tge produced in what may be considered the threatening environment of a classroom.

24 year during the primary grades .From the fourth grade on, results from the Stanford Achievement Test provided information aboutthe.subject' s reading performance. Some of the pertinent findings ofthis massive study are summarized below. 1.Over a period of seven years, the subjects increased the amount of language used in the same controlled situation and increased the smoothness cf their expression. Members of the low subgroup evidenced more difficulty in using and controllinglanguage patterns; they had less to say and had more difficulty in saying it. 2. The subjects varied little in their use of the basic language patterns; the distinction between the high group and the other subjects was defined as greater dexterity in varying elements within the patterns.By Loban's definition, this means such things as the substitution of word groups for singlewords, and the use and placement of movable elements of the pattern. 3.Positive relationships between oral language and written language , oral language and reading, and reading and written language were found. Loban stated that subjects who were good readers at the end of the third grade were those children who had "ranked high in oral language" (p. 69) in kindergarten and the first three elementary school grades.7 In sum, Loban felt that a major purpose of the study was attaineda scientific method for analyzing children's language and for locating certain characteristics that are important for further investigation. The variety of

7The meaning of ranking "high in oral language" is not clarified in the context in which this statement appears. One assumes that Loban was re- ferring to an average of the six teacher ratings previously discussed in this chapter.

25 techniques used in assessing the nature of children'slanguage, the use of structural linguistics in addition to a limited useof transformational grammar on two subjects, andthe longitudinal nature of the work make Loban's study unique in its contribution to knowledge aboutchildren's language. Some weaknesses in this undertaking were inevitable,however. Perhaps in dealing with language assessment it is difficult to use measures that are completely free fromsubjective bias; but the questionable method of determining reading proficiency by the number of booksread could easily have been made more reliable and valid. Standardized reading achievement tests are available for the primary grades; these would have been a more reasonable measure of relationships between certain language characteristics and reading at the primary level. An area that did not receive any consideration was theschool environment.It would be helpful to know, for example, whether or not the teachers were equally qualified in those classrooms that contained students of predominantly good or poor ability. One might ask if children in the low group performed differently from those in the high groupbe- cause they had poorer teachers, or becausethey went to less well- equipped schools.Loban did not discuss these important variables. Thus far in the review, it should be apparent that each of the four researchers (Strickland, 1962; Hocker,1963; Riling, 1965; and Loban, 1963) used a method of oral language analysis derived primarily from structural linguistics which provides, among other things , the description of "arrangements of the patterns of grammatical structure[Fries,1963, p. 73]."In a recent study, O'Donnell, Griffin, and Norris (1967) were influenced more by methods of analysis derived from transformational grammar.In particular, they were concerned with sentence-combining transformations,described in the study as the process of "converting a

26 pair of sentences into a single sentence by embeddingone in the other [p. 35]." The general purpose of theirinvestigation was to analyzesam- ples of speech and writing from elementary schoolchildren so that any differences in grammatical complexities could beidentified and then compared for boys and girls at different gradelevels. Thirty children in each of Grades K-3, 5, and7 comprised the sample; there were an unequal number of boys and girlsat each grade level. All children were pupils in one school in Murfreesboro,Tennessee, with the exception of the kindergartners,who were enrolled in private kindergartens. The language samples consisted ofresponses to two short movies. The procedure was the same for all subjects:First, children viewed the movies, shown without sound, ingroups of threes and were then asked to tell the story to an examiner in private. A series of standardquestions was asked after the child gave his interpretation of themovie. All oral responses were recorded by tape. Second, children in Grades 3, 5, and7 were then asked to write the story of the movie andto write answers to the questions. The purpose of the questionswas to elicit a different kind of response than just narration.It will be remembered from the precedingdiscussion that other researchers handled this problem differently;Hocker (1963) obtained language samples in purposely uncontrolledsituations,while Riling (1965) made every effort to restrict theresponses to narration only. A transcript of each child'sresponses was made, and the typescripts were subsequently segmented into minimal terminableunits (T-units), which have been demonstrated to bemore reliable units of analysis than the phonological units previously usedin studies of this nature.8 Hunt (1965) has discussed this unitas being "the shortest segments which it would be grammatically allowableto write with a capital letter atone end

8Seenote 1.

27 and a period at the other, leavingno fragment as residue." Thus, "The lady with whom I played tennis last nightwas never intended to be and is still not intended to be my wife" isone T-unit, where, "She looked and he looked" is two T-units. Once the samples had been segmented, each T-unitwas typed on an analysis sheet and analyzed according to the followingprocedure:(1) the

"language pattern" of the main clausewas determined (e.g.,noun-verb- object, noun-verb, etc.); (2) sentence-combiningtransformations were classified under three headingstransformations producingnominal constructions, those producing adverbial constructions, andthose producing coordinate constructions within T-units. The data were presented-in terms ofgroup means or rates of occurrence per 100 t-units.Statistical treatment included analysis of variance to test the mean differences in the frequency ofcertain grammatical structures for boys and girls at each grade level and fororal and written language. Among the pertinent conclusions of the studywere the following: 1.Use of the mean word-length of T-units providesa valid and simple measure of the development ofsyntactic control of children. 2.The number of sentence-combining transformationsper T-unit is significantly greater for each advance in grade level,and it is greater in written language in Grades 5 and 7. 3.In the higher grades (5 and 7) control of writtensyntax is far more evident than control of syntax in oral language. 4.Oral expressionprogresses fastest between kindergarten and the end of first grade and again between theend of fifth grade and the end of seventh grade. 5. There is a positive correlation between advancein grade level and increased word-length ofresponse for elementary school children.

28 6. With respect to differences between oral and written language, the length of T-units in oral language was significantly greater than in written language for Grade 3; but in Grades 5 and 7 the written T-units were longer, though not significantly so . 7. No distinct difference between boys and girls was evident in speech. In writing, however, girls in Grades 3 and 5 were superior. The O'Donnell et al. study raises several questions about the "un- impressive progress in syntactic control in the middle grades [p. 100].11 The investigators suggested that research on the production and interpretation of grammatical structures could lead to improved school programs; children's ability to handle structures in their language could be developed for greater flexibility of expression. Besides suggesting several areas for further research, perhaps one of the major contributions of this study is its application, for the first time, of a refined measurethe T-unitto analyses of children's oral language. The results of future studies that are based on this measure should be more dependable than those using the less easily determined phonological unit. Five broad studies of children's oral language have been reviewed for the period 1959-1968 which followed development of a relatively new scheme for analyzing oral language samples. Some of the major contributions of the studies have been noted throughout the preceding review. They can be summarized again briefly as follows: 1. The studies demonstrated that children's language can be analyzed by methods that offer much potential for describing a variety of language characteristics. The two kinds of schemes, onefrom structural linguistics and the other from transformational grammar, make it possible to work from the kindergartner's language to the more complex language of the sixth graders and yet compare the two along the same measures. The studies differed, however, in the amount of attention paid to careful analysis of

29 subordination, or the Level II analysis. The emphasis in Strickland's work (1962) was primarily the categorization of language into language patterns, where Loban (1963) provided a more detailed analysis of the kinds of subordinations used, and O'Donnell, Griffin, and Norris (1967) studied the kinds of sentence-combining transformation used. 2.Of the three researchers who investigated the relationship between reading comprehension and select characteristics of oral language, each provided evidence that some relationship does exist. Children rating high in reading ability seemed to be more flexible in their use of basic language structures (Strickland, 1962; Riling, 1965; Loban, 1963). 3. A previously unnoticed indication of oral language maturity was made possible by the use of transformational grammar, by which means it was discovered that children's use of sentence-combining transformations increases across grades (O'Donnell, Griffin, and Norris, 1967). 4.In three of the five studies,the investigators concluded that the kinds of structures children read in textbooks are unlike the structures in their own speech (Strickland,' 1962; Hocker, 1963; and Riling, 1965). The implication is for research specifically directed to the question of whether or not children would benefit from material that was more like the way they speak. The present study was designed to investigate this particular problem.

Experimental Studies Using Written Material Based on an Analysis of Children's Oral Language Analyses of the language used in materials written for children have been done in an attempt to relate children's use of oral language to the language they read (e.g.,Davis and Seifert, 1967; Strickland, 1962; Riling, 1965). Research on readability has defined various structural

30 elements, like the number of simple sentences or sentence length, that are more or less difficult for children toread.9 The study discussed in this section, however, is the only one, to the investigator's knowledge, that has attempted to experiment with children's materials written according to characteristics of children's oral language patterns. Ruddell (1963) designed a study to test two basic hypotheses: The degree of comprehension with which written passages are read is a function of the similarity of oral patterns of language structure to written patterns of language structure in reading passages Comprehension scores on reading passages utilizing high frequency patterns of oral language structure will be significantly greater than the comprehension scores on reading passages utilizing low frequency patterns of oral language structure [p. 35]. He selected oral language patterns from Strickland's study (1962) which had occurred with either a high or a low degree of frequency at the fourth-grade level, and from them he constructed six 254-word passages, three that used frequent oral language patterns,and three that used infrequent oral language patterns.The patterns appeared in direct proportion to the number of times they were used in the oral language of fourth graders (Strickland, 1962). The variables ofcontent,length, style, and readability were controlled, although the content of the threepassages written

9Summaries of readability research that has dealt with language variables other than the one in this study include those by Gray (1947),Betts (1949), and Ruddell (1963).

31 with frequent oral language patterns was only similar, notidentical, to that of the other three passages. Subjects for the study were randomly selected from the fourthgrade population of Bloomington, Indiana, where Strickland had recently conducted her study of the oral language patterns of elementary schoolchildren (1962). Variables of sex, IQ, mental age, chronological age, and socioeconomic background of the subjects were considered. Reading comprehension of the passages written with frequentand infrequent oral language patterns was ascertained by the cloze procedure, by which every fifth word in the passages was deleted and a blank space left in its place. The reader's task was to fill in the blanks with the deleted words.The tests were scored by two different procedures.In one, the exact same missing word had to be supplied in order for the response to be considered correct; in the second method, a synonym was scored correct. The subjects were randomly assigned to read either the three frequent oral language passages or the three infrequent oral language passages first.By the end of two testing sessions, each subject had read all six passages. Reliability coefficients obtained by the split-half method with odd and even numbers,and then corrected by the Spearman-Brown Formula, ranged from .851-.919 for the individual tests.Ruddell reported that all reliability coefficients were significant at the .01 level. A single measure of validity was obtained by correlating subjects'cloze comprehension scores with scores from the Paragraph Meaning section of the Stanford Achievement Test (Intermediate I, Form XR) . The correlations , which ranged from .609 to .738 on individual tests, were reported to be significant at the .01 level. To test the first hypothesis, a one-way analysis of variance with a repeated measures design was used. Scores from the six passages were

32 treated in the analysis. For both scoring methods, the F ratios were sig- o- nificant at the .01 level. Ruddell concluded that reading comprehension is a function of the similarity of oral and written patterns of language structure. The second hypothesis was also tested by a one-way analysis of variance with a repeated measures design.For this analysis, scores from the passages written with frequent patterns were pooled and scores from passages written with infrequent patterns were pooled. Again, the F ratios were significant at the .01 level for both scoring methods used. Ruddell concluded that reading comprehension of material that utilizes frequent oral language patterns will be significantly greater than reading comprehension of material that utilizes infrequent oral language patterns. The relation of six variables to reading comprehension of the frequent and infrequent passages was determined by using a two-way analysis of variance with a repeated measures design. Significant differences at the .01 level were found in comprehension scores of reading passages that utilized frequent and infrequent oral language patterns and the six variables of (1) occupation of the father, (2) education of the father, (3) education of the mother, (4) IQ, (5) mental age, and (6) chronological age. No significant differences were found between performances of girls and boys.Ruddell indicated, however, that boys had a relatively more difficult time with the passages written with infrequent oral language patterns than did the girls. Although Ruddell was careful to point out that the cloze procedure has been found to be a valid measure of reading comprehension, he did not evaluate its potential bias in a study which dealt with reading comprehension of select language patterns.It seems reasonable to hypothesize that filling in blanks in familiar language structures such as the nounverb object pattern is not as difficult a task as supplying words for language

33 structures that vary the position and typeof certain elements in it.For example, a characteristic of the infrequent oral language patternsin his passages was their use ofmovables, which have been demonstrated to be associated with more mature oral language in children(Loban, 1963). Thus although children might have been able to comprehend asentence that comprised one or more movables , they might have haddifficulty supplying missing words merely because the type of movableand its location were unexpected in that particular sentence. It would have been informative if Ruddell had done an analysisof the location of each child's mistakes in the frequent and infrequentlanguage patterns to see whether there were some relationshipbetween the type of pattern and the mistake made. One might ask, forexample, whether the preponderance of errors in the infrequent patterns occurred within the movable elements rather than in the slots of the sentence.

Summary In the single study that focused on readability in terms of children's oral language, Ruddell (1963) wrote passages composed of either highly frequent or infrequent oral language patterns of fourth graders, asdetermined by Strickland's study (1962). He constructed cloze comprehension tests for each of his six passages and administered them to 100 fourth graders. A major finding was that reading comprehension of the passages that comprised frequent oral language patterns was significantly greater than comprehension of the passages that comprised infrequent oral language patterns. The study was limited to fourth grade, and its use of the cloze procedure to measure reading comprehension could conceivably have affected the results.The findings of this significant work should therefore be confirmed by researchers who investigate the effect of select language patterns by means of a different reading comprehension task, and at different grade levels. The present study was designed to do this.

34 II METHOD

Five areas of the study are discussed in this chapter:(1) overview of the study, (2) development of the reading comprehension tests, (3) subjects, (4) data collection, and (5) data analysis .Estimates of reliability and validity for the instruments devised by the investigator were obtained simultaneously with the results.Consequently, a discussion of these two characteristics of the tests is deferred to Chapter III.

Overview of the Study The general design of .this study was a 2 X 2 X 2 factorial with two treatments (Test A and Test B), two grade levels (Grade 2 and Grade4), and two sexes. Subjects for the study were whole classrooms of second and fourth graders in the two schools made available to the investigator. For each subject, the following data were collected:(1) reading comprehension scores on iest A, which comprised frequent oral language patterns , and Test B, which comprised infrequent oral language patterns;(2) a grade placement score on the Paragraph Meaning section of the Stanford Achieve- ment Test (hereafter referred to as the Stanford subtest), Form Y,Primary II for the second graders and Intermediate I for the fourth graders;(3) a total IQ score from the Lorge-Thorndike IQ test, administered by the schools during the second grade; (4) the occupation of the head of the household; and (5) sex. Variables 3 and 4 were not considered in the

II analyses; they were used only for descriptivepurposes to characterize the sample of the study more fully.

Development of the Reading Comprehension Tests A

To determine whether or not children comprehend materialwritten with frequent oral language patterns better than theycomprehend material written with infrequent oral language patterns,two distinct reading corn- prehension tests were constructed: Test A comprised languagepatterns that appear frequently in the oral language of bothsecond and fourth graders (Strickland, 1962); Test B comprised languagepatterns that appear infrequently in the oral language of both second and fourth graders (Strickland, 1962). In prior research, Ruddell (1963)wrote 254-word passages which used select oral language patterns in direct proportionto the number of times they appeared in children's speech.' Theuse of a prescribed number of language patterns in eachpassage resulted in material that was some- times stilted and contributed.2To avoid this problem in the present study, reading comprehensionwas ascertained for individual sentences thatwere unrelated to each other; the proportion ofeach language pattern used in the tests, however, was still controlled.The decisionto use unrelated sentences rested on three assumptions:

4 1This study was reviewed in ChapterI. 2 To clarify this point, these samplesentences are quoted respectively from Ruddell's Passage IA and IIB:(1) "Man must learnmany new things about space and he plans futurespace flights"(p. 139), and (2) "After catching fish a Japanese fisherman sells froma big water tub and while in the water the fish will live" (p. 146).

36 mr.

1. In a lengthy.passage it is difficult to check the degree of comprehension of each sentence. When working with language patterns that differ in frequency, the variable under careful scrutiny in this study, it is desirable to measure reading comprehension of each sentence. 2.Reading a series of unrelated sentences rather than a lengthy passage, such as those used by Ruddell (1963), was assumed to be better suited to the abilities of the younger subjects in the study. 31 Although second and fourth graders typically read short stories or connected paragraphs rather than isolated sentences,comprehension of a unit as small as a sentence is frequently required in the classroom. On the basis of these considerations,the tests were a series of sentences.Briefly, three pictures were drawn for each sentence. The reader's task wasp to select the one picture that "told about" the sentence the best and then draw a line from the sentence to the picture. The procedure for selecting the patterns and exerting control over extraneous variables is outlined below; a more complete discussion of how reading comprehension was measured by multiple choice items appears later in the chapter.

Selection of the Patterns There have been several recent descriptions of the structures that appear in children's orallanguage,3 but fortwo reasons only Strickland's findings (1962) were used.First, Loban (1963) defined nine patterns into which all oral communication units can be classified .4 He determined the

3Five recent studies that focused on such descriptions were reviewed in Chapter I. 4 Some examples are the subjectverbdirect object pattern (124) and the subjectverbindirect objectobject pattern (1234).See the review of Loban's study in Chapter I for a complete list of these patterns.

37 degree to which each one appeared at various stages of language development of elementary school children. However, because widely different variants of each pattern are present in children's language (Loban, 1963;Strickland, 1962; Riling,1965), the present studywas not confined to Loban' s list. Second, in similar studies,Riling (1965) and Hocker (1963) used Strickland's work (1962) as the impetus for their research. Theywere not as comprehensive; the sample for the former comprised fourth andsixth graders while for the latter it comprised only first graders.Strickland's study (1962), then, with its broad description of structures inchildren's oral language was the sole source of information about the nature andfrequency of children's oral language patterns. One of the major interests in the present studywas the comparison of reading comprehension scores of second andfourth graders. Two alterna- tives were available in constructing thematerials: either write a different set of materials for each grade, or writea set of materials for second graders and give them to subjects at both gradelevels. The first alternativewas discarded because obtained grade leveldifferences in comprehension of select oral language patterns would have beenconfounded by all the factors that differentiated the tests at two grade levels.The second alternative was accepted; control over variables ofcontent, length, grammatical complexity, and vocabularycontrol thatwas essential if the focus was to be language pattern difficultycould be identical for allsubjects .5

5 In a recent study in which the quality ofmain idea statements made by elementary school children in Grades 1-6were compared (Barrett and Otto, 1968), materials were constructedin two ways:(1) a short paragraphwas written at the first grade level and givento half the subjects across grades and (2) the paragraph was gradually mademore difficult so that the readability level increased for each succeeding gradelevel. The authors reported significant differences in the main ideastatements of succeeding grade levels regardless of whetheror not the first grade or the appropriate grade-levelpar- agraph was read by the subjects.From this finding one could reasonably infer that the main idea taskwas not adversely affected by having older subjects read the presumably easier first gradeparagraph. This finding is considered relevant to the decision to give thesame tests across grades in this study. 38 Because the same materials were tobe used, a major problem was selecting language patterns fromStrickland's study (19 62) that appeared in approximately the same rankorder of frequency in both second and fourth grades. This was necessary to assurethat the meaning of "frequent" and "infrequent" was the same forsubjects at both grade levels. This consideration superseded all others until therequirement was met. Patterns for Test A were selected from amongthose that ranked the highest in frequency at the second and fourthgrades in the Strickland study (19 62).Patterns that were ranked lower than tenthin frequency appeared less than one-sixth the number oftimes the most frequent pattern at either grade level appeared; thusit seemed reasonable to designate as infrequent any pattern lower than tenth in therank order. The- list from which the patterns for each test were drawn ispresented in Appendix A. Three additional criteria were considered indetermining which patterns were to be used: 1. Any pattern that ended with a T marker, aconjunction indicating that another pattern should follow (e.g. , 1 2 4 +T), had to be followed by a pattern beginning with a T marker, aconjunction indicating that another pattern should precede it (e.g. , T 12 4). 2. Since two of the most frequent patternsdid not contain a T marker, or more than one main clause(1 2 4 and 1 2b 5),an attempt was made to include a similar number of infrequent patterns withonly one main clause so that neither testwould be overburdened with lengthy sentences.Be- cause it is generally agreedthat sentence length influences readability, control of this variable seemed necessary. -3.Patterns, were to appear in approximate proportion to the numberof times they appeared in children's oral language. The patterns selected for Test A are presented inTable 1 and patterns selected for Test B are presented in Table 2 alongwith information about the rank order and frequency for each grade level. As shownin the tables,

39 Table 1 The Rank Order and Frequency of Oral Language Patterns Selected for Test A

Oral Grade 2 Grade 4 Language Rank Rank Patterns Order Frequencya Order Frequencyb

1 2 4 1 309 1 202 T 12 4 2 264 2 156 12 4 + T 3 189 3 123 12b 5 4 130 4 108 T 12 4 + T 6 74 5 84 a These figures are taken fromdata on the 100 second graders in Strickland's study (1962). b These figures are taken fromthe data on the 75 fourth graders in Strickland's study (1962). several differences between the patterns of Test A and Test Bare apparent. First, the rank orders of the infrequent patternswere less similar across grades than the frequent patterns.Since the difference in rank order was never greater than two, however, the discrepancy was considered insig- nificant for purposes of this study. A second characteristic of the infrequent patterns was their diversity, which contrasted sharplyto the 12 4 component in all but one of the frequent patterns.This diversity was expected; research has indicated that flexibleuse of the basic language structures is a trait of mature and thereforemore individual oral language in children (Loban, 1963). A third difference between the two testswas the larger number of patterns selected for Test B; five were used in Test A compared to sixin

40 Table 2 The Rank Order and Frequency of Oral Langua4e Patterns Selected for Test B

Oral Grade 2 Grade 4 Language Rank Rank Patterns Order Frequencya Order Frequencyb

T 12 M2 17 32 17 25

12 M2 17 32 19 22

T 12 4 M2 19 29 20 21 M3 12 Mi +T 20 28 21 19

12 34 23 23 24 15

12 4 M, 25 21 23 17 aThese figures are taken from data on the 100 second graders in Strickland's study (1962). These figures are taken from data on the 75 fourth, graders in Strickland's study (1962).

Test B. Adding a sixth to Test A would have meant` including a pattern of much lower frequency of occurrence than the five top-ranking patterns that were used. A pattern with a T marker was added to 'Test B-- making two patterns that ended with a conjunction and two that began withoneso that an approximate similarity of the oral and written language could be maintained.

Determination of the Proportion of Patterns. to be Used ,Children's oral language patterns differ greatly in the frequency of their occurrence. To assuresome similarity between the structure of

41 children's oral language and the written languageof the tests,each pattern was used in approximate proportion to the numberof times it appears in children's oral language (Strickland, 1962). Several steps were necessary to determinehow many times each pattern should be used. As demonstrated previouslyin Tables 1 and 2, the frequencies and rank orders of thepatterns differed from grade to grade. The general procedure was tocompute the proportions separately for each grade and then average them.More specifically, these computations were made for each test: 1. The frequencies of all the patternswithin the test were added. This process was repeated separatelyfor Grades 2 and 4. 2. The frequency of each patternwas then divided into the resulting sum, yielding an index for eachpattern. 3. To find the average index for eachpattern across Grades 2 and 4, the second and fourth grade indiceswere added and divided by two. This final number was used to determinehow many times each pattern would appear. An example is provided for clarificationof this process: In Test A, the frequencies of the fiVe patternstotaled 966 at the second grade level. This total was divided into the frequencyof the 12 4 pattern, 309, to determine its index. The resultingnumber, .32, was the index,or proportion, for the 12 4 pattern. Derived in the samemanner, the index for the 12 4 pattern at the fourth grade level was .30. Averaging.32 and .30 resulted in the final index, .31. Thus 31% of the frequentpatterns in Test A should have been 1 2 4 (nounverbobject). The indices computed for eachpattern in Test A and Test Bare summarized respectively in Tables 3 and4 along with the actual proportions used.

42 --77T7'73

Table 3 Index of Proportions of Frequent Oral Language Patterns in Material Written for Second and Fourth Graders

Index Frequent (The average between Oral Language proportions determined Actual Pro- Patterns separately for Grades 2 and 4) portions Used

1 24 .31 .32 T 1 24 .25 .23 1 24 +T .19 .23 1 2b 5 .15 .11 T 1 24 +T .10 .12

Table 4 Index of Proportions of Infrequent Oral Language Patterns in Material Written for Second and Fourth Graders

Index Infrequent (The average between Oral Language proportions determined Actual Pro- Patterns separately for Grades 2 and 4) portions Used

T 1.2 M2 .20 .15 1 2M2 .19' .18 T 1 2 4 M2 .18 .20 M3 1 2Mi +T .17 .18 1 2 3 4 .14 .15 1 24 M1 +T .14 .16

43 Control of Select Variables

Since the primary focus of the studywas to determine whether or not reading comprehension of the patterns inone test differed from reading comprehension of the patterns in the other test, controlover vocabulary, content, and grammatical complexitywas essential in both tests.Fur- thermore, since the same testswere given to two grades,it was necessary to demonstrate that the three variables were controlledacross grades so that the research task did not inadvertently bias the performance of one grade over the other. The manner in which the influence of each variable was defined and controlled is discussedbelow.

Vocabulary control. Because thesame tests were read by children in second and fourth grades, the intentwas to make the actual decoding process as uncomplicated as possible for the formerso that performance did not depend on familiarity with thewords. Two types of controlswere imposed. First, the primarysource of vocabulary for both tests was the Stone List (Stone, 1953),a revision of the Dale List of 769 Easy Words. Words that did not appearon the list were given to second and fourth graders in a pilot study;6 ifa clearly larger number of the former misread a word, then a Stone List word was substituted in the finalversion. A second kind of controlwas inherent in the fact that the subjects at both grades were directed to raise their handif they needed help withany words they did not know.(See Appendix B fora copy of the Directions .) From pilot studies that preceded final testing, itwas clear that second

6 The words were listed in random order.In a separate testingroom, children from two second grade classroomsread the words to theexam- - finer.Mistakes were recordedon a separate list,

44 graders in particular relied on and welcomed help with words if they needed it.Even the simplest words were requested by.some of the childrenwords that appeared on the Stone List. Though it was not practical to use thesame words in each test,7 an attempt was made to overlap the vocabularyas much as possible. Thus two additional constraints were imposed:(1) the same number of "hard" words (words not appearingon the Stone List) were used in both tests; and (2) a type-token ratio which has been used inprior research (Loban, 1963; Kean, 1967), was madeas similar in each test as it was possible and practical to do. In this ratio, the number of different words used (type)is related to the total number of words used (token). In Table 5 the "hard" words for each testare listed, and in Table 6 the type-token ratios for Test A and TestB are presented.

Content control. Because the vocabularywas drawn largely from the Stone List (Stone, 1953), the content of eachtest was limited to objects and actions that were within the readingvocabulary of a typical second grader. To attempt to write reasonablyinteresting test materials for both grades,the following criteriawere used to determine the content of sentences within the tests: 1. Highly implausible situations whichrequired knowledge of difficult relationships of factswere avoided in order to make the concepts and ideas familiar to the second and fourthgraders alike.

7 The to be verb, for example, appearedin one pattern in Test A (12b 5) and in no patterns in TestB. The verbs were, are, and is followed by the predicate nominativepermeated portions of Test A whereas they were not included in Test B.Thus it was not possible to match the vocabulary of somethingas apparently straightforward and simple as the verbs.

45 Table 5 "Hard" Words Usedin Test A andTest B

Test Hard Words Test A (frequent broken bone bite dried oral language goes kite larger mailbox patterns) longest playful smaller Test B (infrequent airport cream closely drums oral language goes glass lying melts patterns) quickly boots threw

Table 6 Ratio of the Numberof pifferent Words(Type) to the Total Number ofWords (Token) inTest A and Test B Test Type-Token Ratio Test A (frequent oral 182 (type) language patterns) .47 387 (token)

Test. B (infrequent oral 204 (type) .43 language patterns) 469- (token)

2. Closely relatedto the first criterionwas the use of situations that involved animalsand family activities in most ofthe sentences.In

46 materials like basal readers,these topics are widely accepted as being of interest to children at the grade levels used in this study.It was assumed that they provided areas of common experience so that the younger subjects were not put at a disadvantage. The content of both tests was made as similar as possible, but the obvious possibility of using identical situations was avoided for two reasons.First, the tests were given a day apart from each other; memory was expected to interfere with responses made to the second test. Sec- ond, the movable elements that characterized the infrequent patterns in Test B concerned details of time, manner, and placeconcepts that were not a part of the frequent patterns of Test A. Sentences in Test B, then, frequently involved details illustrated by the movables "above the tree," "now," and "near the door" that could not be incorporated logically into sentences in Test To provide some measure of consistency between the tests, however, the content of each test was categorized. The number of sentences falling into eight groups was tabulated and compared with the number in the other test; unduly large discrepancies between the tests were corrected. The number of sentences falling into the eight categories according to sentence content is summarized in Table 7 for each test. The number of sentences in each category according to picture content is summarized in Table 8. On the basis of the tables, these observations can be made:(1) the two tests dealt with similar content as determined by the written sentences in each test, and (2) the two tests dealt with similar content as determined by the picture content in each test.

Grammatical complexity control.Strickland's study of children's oral language patterns (1962) focused primarily on the type of slots in the patterns (e.g.,subject, verb, direct object) and the type and location of

47 Table 7 Analysis of Sentence Content in Test A and Test B

Category Number of Sentences Fallinin Cate or Test A Test B

Animals I3 7 Children (male and female) 4 5 Males 5 8 -Females 2 3 Toys and objects 9 6 Nature 2 4 Animals and males 2 2 Animals and females 2

Table 8 Analysis of Picture Content in Test A and Test B

. Category. Number of Picture Sets in Each Category* Test A Test B

Animals 13 7 Children (male and female) 4 7 Males 5 10 Females 3 2 Toys and objects 8 3 Nature 2 4 Animals and males 2 2 Animals and females - 2

*A picture set is defined as the three pictures that comprisedthe multiple choice items for each sentence.

48 the movables.How each slot is filled, however, seems to be of crucial importance in distinguishing mature oral language (Loban, 1963). Edu- cational experience and reason readily lead to the conclusion that for children the reading task is complicated by putting phrases and clauses in slots instead of single words. Given the wide range of possibilities for each slot, the decision was made to include different kinds of fillers (e.g., single words, phrases, clauses) rather than limit the selection to single words. On the other hand, it was necessary to avoid overloading the slots with complexities that would bias the performance of the younger subjects. To balance these two concerns, the 'preponderance of slots were filled with single words, words modified by one or two adjectives,and compound nouns, verbs , and objects. Phrases and clauses were used more sparingly with the exception of constructions in movables, manyof which were filled with a phrase to avoid needless repetition of single-word adverbs."In winter" and "at night" are examples of substitutions for single-word adverbs in the infrequent oral language patterns that contained the movable M3the adverb of time. An attempt was made to distribute the kinds of slot fillers among the various patterns within each test. This precaution was intended to avoid overloading the simplest and shortest patterns within eaCitte-st-so that they became proportionally more difficult thansome of the longer patterns. Indices of subordination from prior research (Loban, 1963; Kean, 1967) did not readily lend themselves to distinguishing among the kinds of slot fillers controlled in the present study. For thepurpose of convenient classification, the following indexwas devised. Examples for each of the six categories are provided:

49 Category Slot Filler Example for the Subject Slot 0 single word The boy ran.

1 single word + adj ective The slow boy- ran. single word. + adj ectives The slow, tired boy ran. 2 phrase The boy with the coat ran. phrase + adjective The boy with the red coat ran. phrase + adjectives The boy with the torn red coat ran. 3 compound The boy and girl ran. compound + adjective The boy and tall girl ran. compound + adj ectives The short boy and tall girl ran. 4 clause The boy who had a coat ran. clause + adj ective The boy who had a red coat ran. clause The boy who had a torn red + adj ectives coat ran.

It is clear that for other purposes, the index would need to be revised; but it was adequate for its limited use in describing the kinds of grammatical complexities utilized in the present study. The Index Rating (IR) given to slot fillers in Test A is summarized in Table 9 for each pattern, and the IR for slots in Test B is likewise summarized by pattern in Table 10. From inspection of the subject slot for the 1 2 4 pattern in Table 9, for example, 8 subject slots were filled with single words, 5 with single words + adjective(s), 3 with phrases, 1 with a compound subject, and 1 with a negative. To make comparisons between the tests more readily apparent, only the slots that patterns in both tests had in common (subject,verb, object), were analyzed; consequently movables and the indirect object of certain

50 Table 9 Total Index Ratings (IR) Given to Slots in the Oral Language Patterns of Test A.

Frequent Subject Verb Object Predicate Oral (1) (a. or f2b) (4) Nominative (5) Language Patterns IR IRr,. IR IR in Test A 0 1 2 3 4 03 \-5- 0 2 3 4 0 1 234

1 2 4 8,5,3,1, 1 16,1, 2 8,6,1,3,0 T 1 2 4 8,6,0,0, 1 13,0, 1 8,6,1,0,0 1 2 4+T 8,6,1,0:0 11,2, 1 8,6,1,0,0 12b 5 3,2,2,0, 0 7,0, 0 2,2,0,2,0 T 1 2 4+T 6,1,1,0, 0 7,1, 0 4,4,0,0,0 patterns in Test B were not included. A summary of the analysis for each zest is presented in Table 11, where it is clear that the slot fillers were similar from test to test.It was concluded that the grammatical complexities were adequately controlled across tests.

Measurement of Reading Comprehension Since comprehension of each sentence in the tests was assessed, more typical kinds of tasks such as selecting the correct sequence of events or a title for a story were not appropriate. Pictures have been used to measure reading comprehension in standardized reading achievement tests (Gates Advance Primary Reading Test, Type APR, Form I) and in recent research on sentence constructions (Ruddell, 1965). In the latter instance, four sentences were listed opposite two pictures. The task was to'draw a line from each picture to the one sentence that told about the picture.

51 Table 10 OralInfrequent of TimeAdverbTotal Index Ratings (IR) GiVen Subject Verb to Slots in the Oral Language Patterns IndirectObject Adverbof Test B Adverb PatternsLanguage , (M3) (1) (2) (3)* Object (4) of Place (M1) of Manner . (M2) TestBin 0 1 IR 2 3 4 0 1 IR 2 3 4 0 3 5 IR 0 1 IR 2 3 4 0 1 IR 2 3 4 0 1 IR 2 3 4 . 0 1 IR 2 3 4 12341 2 M2 4,4,1,0,1 9,1.,0 1,1,4,3,0 4,3,2,0,0 7,2,0 7,2,0,0,0 , 2,3,1,2,1 M312M1 8,0,4,0,1 7,411,1,0 13,0,0 ,-, 2,0,11,0,0 T12M2T124M124M 1 2 7,2,0,9,07,4,1,0,07,3,0,0,0 11,1,0 7,1,17,2,1 8,3,1,0,06,4,0,0,0 3,0,7,0,0 2,1,6,0,03,1,5,0,1 Table 11 Summary of Total Index Ratings (IR) Given to Slots in Oral Language Patterns Common to Test A and Test B

Subject Verb Object (1) (2) or (2b) (4) T es IR IR IR 0 12 3 4 0 3 5 0 1 2 3 4

Test A (frequent oral language patterns) 33,20,7,1,2 54,4,3 28,22,3,3,0 (N = 63) (N = 63) (N = 56) Test B (infrequent oral language patterns) 36,20,5,1,1 54,7,2 16,10,2, 2,1 = 64) = 64) (N = 31) NoteThe numbers in the slots differ according to thenature of the pattern being analyzed. In the 1 2b 5 patternfor Test A, for example, only the first two slots were includedfor analysis here.

Considering both the relatively large number of sentences tobe read in the present study-37 in eachtest8and the endurance limitations of the younger subjects, it seemed reasonable to require one sentencerather than two to be read for each response. For each sentence, then,three pictures were drawn. One depicted the content of the sentenceand was the correct response. Each of the other two picturesheld most of the details of the correct picture constant while varying one majordetail at a time.

8The reason for this particular number of items is discussedlater in the chapter under the heading, "Final Selection of the Test Items ."

53 An example of the natrtze of the distractorsis provided in Figure 1; the.item is from Test B.It was intended that the basic likeness ofthe three pictures In each multiple choiceset would require careful 'reading of the sentence in order to makea correct response. Reading comprehension in this study, then,was measured by the child's ability to reada sentence and select one of three similar picturesthat belonged with the sentence. The artist was directed touse simple line drawings and to make each of the three pictures similarexcept for the major change in each distractor.If a boy and girl were to be thesubject for a set of pictures, for example, thesame boy and girl were used in each picture although their poses might be differentdepending on the discriminationbeing required of the reader. Figure 1illustrates this criterion. A rigid formula for constructingeach distractorwas not appropriate, as variety in the content and grammaticalcomplexities necessitated flexible treatment of thesentence. Certain guidelineswere established, however, to assure that (1) differentkinds of discriminationswere included within a test, and (2)the discriminations usedin one test were similar to those used in the othertest. An attemptwas made to empha- size reading comprehension ofthe particular characteristicof the pattern being read. In the sentence"Sue does give Tornsome fish" (the infrequent pattern 12 3 4), for example,one of the picturei showed Tom giving Suc some fish whilea second picture showed Sue givingTom one fish. The following types ofdiscriminations were used:

Variable Correct Picture One Distractor (1) noun-verb The dog sees the The dogs see the cat. agreement cat,. (2) negative- Mr. Brown's new Mr. Brown's coatcomes affirmative coat comes with- with a pocket. out a pocket.

54 The young boy without a coat hits with a bat. Fig. 1'. Sample Item from Test B. Variable Correct Picture One Distractor (3) count nouns Just Jane's kite Two kites have two tails. has two tails. (4) mass nouns The fish are eat- The fist' is eating the the food, food. (5) verb tense The boy in the The boy in the coat has coat will throw thrown the ball. the ball. (6) manner* She draws quickly. She draws slowly. (7) place* My friend draws a My friend draws a house house on the paper. on the sidewalk. *These discriminations,which involved movables, wereonly relevant for some of the infrequentoral languaae patterns used in Test B.

Final Selection of Test Items To obtain some feedback from children and also statistical information about the tests and their items prior to final testing,a pilot study was conducted. Two whole classrooms of second graders andone of fourth graders, the number of classrooms made available ina nearby elementary school, served as subjects. From an inspection of several basal readersat the second grade level, a reasonable length for each of the final testswas judged to be 'approximately 35'-40 itemstheaverage number of sentences in the second grade stories. About one and one-half thatmany items for each test, distributed proportionallyamong the patterns within each test, were constructed and piloted. The frequent and infrequent oral language patterns were grouped separately into two booklets, whichwere preceded by a brief set of directions and a practice sentence. Asa check on the clarity and unambiguity of the pictures,subjects were directed to raise their hands if they did not understand a picture. They could also ask for any words that they did not know.

56 1

The children were randomly assigned to read eitherTest A (frequent oral language patterns) or Test B (infrequent orallanguage patterns) first. A test was read on each of two consecutivemornings of testing for each classroom. The data for each subject compriseda single score for each of the reading comprehension tests.In addition, to help determine whetheror not an item discriminated between good andpoor readers, arating of from 1 to 3 given by the teacheron the basis of reading achievementwas obtained. In analyzing the data,a distinction. between grade levelswas not made due to the small number offourth graders. Bymeans of the 1604 computerprogram FORTAP, a Fortran Test Analysis Packageby Baker and Martin (19 68) at theUniversity of Wisconsin ComputingCenter, an item analysis was performed.Among the information provided by this analysis were the following:item difficulty (P),item correlation with the rest of the test (R), X50,and Beta. In addition,a Hoyt Reliability Coefficient was obtained foreach test, the reliability coefficientbeing .85 for Test A and .86 for TestB. A decision was made to weighthe following factors in determining which items were to be consideredfor the final study:(1) item difficulty, as expressed by the proportion of subjectswho responded correctlyto the item (P);(2) correlation of theitem with the rest of t!-1,-1test (It) ;(3) the degree to which the itemdiscriminated between good andpoor readers, as determined by the teacher ratings foreach subject; (4) the degreeto which it was felt that replacingone of the distractors would improve the discriminating power of the itemconsiderably. An example ofa distractor that functioned poorly in thepilot study is provided in Figure2; the item, from Test B, was modifiedso that the middle distractor was replaced by a prezrumably less obviousone. The revision is presented in Figure 3.

57 She buys her children some milk and eggs. Fig. 2. PilotA Poor Study Distractor from Test(The in the B. Middle Picture) She is buying her own children some milk and eggs. Fig. 3. The Revised Distractor (The Middle Picture) from Test B. When the best items were selected according to the above criteria, an attempt was made to improve them so that the tests would discriminate more effectively between the good and poor readers. The mean scores for both tests were considered higher than desirable, being 47.66 for Test A and 43.33 for Test B out of a possible 57. Thus, to make the tests more difficult, one or both of these procedures was used:(1) the distractors were made more difficult by requiring finer discriminations among the three pictures for each sentence, and/or (2) the sentences were made more difficult by expanding them slightly in terms of length, the addition of adjectives, or the addition of more complex grammatical structures like phrases. An example of the former was provided in Figures 2 and 3. An example of the latter procedure is in the sentence in the Test A pilot study, "The monkey has a balloon and a toy car." As it appeared in the final version, the sentence read, "The monkey went and gota balloon and a toy car."- Accordingly, some of the items were revised. A final selectionwas determined by the proportion that each pattern was to appear in the test.9 The final number of items that met these requirements was 37 in one test and 42 in the other. Items were deleted from the latter until both tests contained a total of 37 sentences. See Appendix C for a list of the 37 sentences in each test, analyzed in terms of the language patterns used.

Composition of the Final Tests In sum, each of the two final versions of the reading comprehension tests comprised 37 sentences --two on each page but the last, which

9 For a discussion of how the proportion of each pattern was determined, refer to pp. 41-43.

60 contained one sentence. Within each test, the 37 sentences were randomly ordered; and for each sentence the three pictures were randomly ordered. See Appendix B for a copy of Test A and Test B. A set of directions preceded each 18-page test. The directions for the first day of testing were somewhat longer than for the second day. During each, day, the subjects were told to read a sentence and look carefully at the three pictures under the sentence. They were to decide which picture told about the sentence the best. Then a line from the sentence to the picture was to be drawn. They could raise their hands and ask for any word that they did not know. There was no time limit imposed. In addition to the directions there were two practice sentences for the first day and one for the second day. These items were devised so that the patterns were neither the frequent nor the infrequent patterns used in the tests. A copy of the Directions for each day of testing is in Appendix B.

Subjects The subjects were from two elementary schools in Madison, Wisconsin. The schools were chosen because the students were presumed to be reasonably similar with respect to socioeconomic background and spread of ability. School I had five second-grade and four fourth-grade classrooms, and School II had two classrooms at each grade level. All second and fourth graders, within these 13 classrooms were designated subjects.The exceptions were children who had repeated a grade. This included 20 second graders and 13 fourth graders. In addition, a teacher asked that one child not be required to participate because of emotional instability. A final total of 163 second graders, 81 girls and 82 boys, and 137 fourth graders, 69 girls and 68-boys, served as subjects. In order to characterize the sample population, IQ scores and a measure of, socioeconomic background were obtained for each subject. Results

61 from the Lorge-ThorndikeIQ tests administered bythe schools in second grade were used. The mean, range,and standard deviation of IQ scores for all subjects and forsubjects grouped by grade, sex,and sex within grade are summarized inTable 12. In the populationsample, ability ranged from very low (an IQof 77) to very high (an IQ of145); a wide range of ability wasalso evident in each of thesubgroups. The mean IQ of the sample was 105.6, orslightly above average. As thelargest difference between the subgroup meansand 105.6 was 1.6 (105.6 minus the fourth-grade boys' meanof 104.0), it was concluded thatfor the sample as a whole and for eachsubgroup, the subjects could bedescribed as approximately"average" in terms of their meanIQ scores.

Table 12 The Mean, Range, andStandard Deviation of IQ Scores for all Subjects and for SubjectsGrouped by Grade, Sex, and Sex within Grade

Group Mean Range Standard Deviation

All Subjects 105.6 76-145 11.7 Second Graders 105.9 82-134 11.2 Fourth Graders 105.3 76-145 12.2 All Boys 104.8 76-145 13.0 All Girls 106.3 77-136 13.6 Second Grade Boys 105.6 84-130 11.8 Second Grade Girls 106.1 82-124 9.5 Fourth Grade Boys 104.0 76-145 12.7 Fourth Grade Girls 106.5 77-136 11.2

The Duncan Socio-economicIndex Scale was used to depictthe subjects' socioeconomic background.As described by Reiss(1961), a

62 subject's position on the scale is determined by the occupation ofthe head of thehousehold.10 Each subject was assigned a score from the index according to the parent's occupation. Then, using Mortenson's method (1966), the occupational levels were combined to yield three socioeconomic levels: a rating of 0-2 was considered low, 3-6 was considered middle, and 7-9 was considered high. The proportion of subjects in each of the three categories is summarized in Table 13. As shown, the spread of occupational levels ranged from the lowest to the highest cate- goriesbut over half the sample was rated at the middle socioeconomic level.

Table 13 Proportion of Subjects in Three Levels Defined by the Duncan Socio-economic Index Scale

Index Rating 0-2 3-6 7-9

Proportion of Subjects .26 .56 .18

In sum, the subjects could be characterized as approximately "average" in terms of their mean IQ score, and the majority could be described as being from the middle socioeconomic level.

10The validity of this measure was demonstrated by Reiss (19 61) when he correlated scores from the Duncan scale with ratings of income and education. The rank correlations between the Duncan scores and income was .85 and between the Duncan scores and education was .83. Reliability of the Duncan scale has not been determined, although a product moment correlation coefficient of .99 was obtained between 1947 scores of the scale from which the Duncan scale was derived and a replication in 1963 (Hodge, Siegel, and Rossi, 1964).

63 Data Collection Pupils in each of the sevensecond grade and six fourthgrade classrooms weretested by classroom for three separatetesting sessions. Dur- ing one week of April,1968, the reading comprehensiontests were administered on two consecutivemornings to subjects in eachof the thirteen classrooms. During the followingweek the Stanford subtest wasadmin- istered. Primary II to the secondgraders and Intermediate I tothe fourth graders. This subtest consists of aseries of paragraphswhich are grad- uated in difficulty. From eachparagraph, one or two words aredeleted. The reader's task is to seU,ctthe correct word for eachblank space from among four choices.The timed test is 30 minuteslong for second and fourth grades.11 To avoid bias at either gradelevel, the second- and fourth-grade classrooms were evenly distributedthroughout the three testing periods each day. Thus, some second and somefourth graders were tested at either 8:30, 9:30, or 10:30 in themorning. The reading comprehension tests were administered at the sametime of day to any one classfor each of the two days. Although anhour was allowed per class, theslowest readers took approximately 35 minutes. All of the testing war done by threeexperienced elementary school teachers who became familiar with theintent and procedures of thestudy

11According to the test manuals for PrimaryII and Intermediate I (Kelley, Madden, Gardner, and Rudman,1964), the Kuder-Richardson reliability coefficients and standard errorsof measurement for each test are, respectively, .93and 2.0 at the second grade leveland .91 and 4.0 at the fourth grade level;these measures were based on arandom sample of 1000 pupils at each grade.The authors discussed content validity in terms of their examinationof courses and textbooks todeter- mine the nature of theunderstandings to be measured.

64 TI

in a training session prior to testing. Within the two schools, each examiner was randomly assigned to a block of both second- and fourth-grade classrooms; only two examiners were needed to complete the testing for School II, which had four participating classrooms.Subjects who had been absent for one of the testing sessions were tested in smallgroups during a third week.

Data Analysis To provide answers to the three basic questions of the study,' 2 nine hypotheses were tested: Hl:The number of second and fourth graders whose reading comprehension of material written with frequent oral language patterns is better than reading comprehension of material written with infrequent oral language patterns is not significantly greater than the number of second and fourth graders whose reading comprehension of material written with frequent oral language patterns is not better than reading comprehenSion of material written with infrequent oral language patterns. H2: Fourth graders do not do significantly better_ thansecond graders in their reading comprehension of material written with frequent oral language patterns. H3:Fourth graders do not do significantly better thansecond graders in their reading comprehension ofm_ aterial written with infrequent oral language patterns. H4: Second- and fourth-grade girls do not do significantly better than second- and fourth-grade boys in their reading comprehension of material written with frequent oral language patterns. H5: Second- and fourth-grade girls do not do significantly better than second- and fourth-grade boys in their

12Refer to page 4 of Chapter I for a list of the questions.

65 reading comprehension of materialwritten with infrequent oral language patterns. 116: Second-grade girls do not dosignificantly better than second-grade boys in theirreading comprehension of material written with frequentoral language patterns. 1-17: Second-grade girls do not dosignificantly better than second-grade boys in theirreading comprehension of materials written with infrequentoral language patterns. H8: Fourth-grade girls do not dosignificantly better than fourth-grade boys in theirreading comprehension of material written with frequentoral language patterns. 119: Fourth-grade girlsdo not do significantlybetter than fourth-grade boys in theirreading comprehension of material written withinfrequent oral languagepatterns. The manner in whichthe hypotheseswere tested is discussed below.

Hypothesis 1

Byuse of the chi square test,one can determine whethera set of observed frequencies isconsistent with the frequenciesthat are expected -if the hypothesis is accepted (Tate, 1955).If the discrepancybetween the observed and expectedfrequencies is too largeto be credited to sampling fluctuations,the hypothesis is rejected.With respect to the hypotheiis underconsideration, chisquare was used to determine whether or not the number of subjects whocontained a higher readingcomprehension score on Test A (frequentoral language patterns)than Test B (infrequent oral languagepatterns)was significantly greater thanthe expected number.

Theoretical expectationwas determined by dividing thenumber of subjects in each chisquare analysis by two. Thisexpectation of N/2 provided for a 50-50chance thatscores would be higheron one test than on another.If there wereno discrepancy between thistheoretical frequency and the observedfrequency, then the valueof chi square wouldbe zero.

66 The statistic used was the following, where fo= the observed frequency and fe = the expected frequency: (fo f )2

Hypotheses 2 and 3 For Hypotheses 2 and 3, the t test was used to test the significance of any observed difference in meanscores between the performance of second and fourth graders on the reading comprehensiontests. Since the t test is based on the assumption of equal populationvariances, preliminary tests were carried out to providea check on whether or no-t' this was the case; the F ratio was computed for eachhypothesis being tested. When the obtained value of Fwas significant at the .05 level, the following model for the t testwas used (Winer, 1962):

t= (Ea (11a 2 Sa+ Sb2 Na Nb

When the population variances could be assumedequal, this model for the t test was used (Hays, 1g63):

(M1 -M2) t= -E(m1 -m2) (N iS?+ N 2S22) N 1 +N 2

(N1 + N2-2) N 1 N 2

67 Hypothese s 4-9 For Hypotheses 4-9, the t test was used to test the significance of any observed difference in mean scores between the performance of boys and girls on the reading comprehension tests, both within and across grade level. To check the assumption of equal variance on which the t test is based, an F ratio was computed for each of the two groups being compared.If the obtained value of F was significant at the .05 level, the model for the t test which assumes unequal population variance was used as previously described for Hypotheses 2 and 3. When the population variances could be assumed equal, a different model for the t test, also described for Hypotheses 2 and 3, was used.

Summary of Analysis Hypothesis 1 was tested by means of chi square. For the remaining eight hypotheses, oneof two models for the t test was used depending on whether or not the population variances could be assumed equal.

Summary In this chapter a general overview of the study precededa discussion of how the investigator devised the instruments to measure children's reading comprehension of select oral language patterns. Procedures forcon- trolling variables like content and grammatical complexity from test to test were outlined, and criteria for selecting the final items were presented. The chapter also contained information about the subjects, whowere characterized as "average" in terms of IQ and socioeconomic background. Testing procedures were discussed and the statistics used to test the nine hypotheses were outlined.

68 III RESULTS AND DISCUSSION

In the present study therelationship between readingcomprehension and material written withfrequent and infrequentoral language patterns was investigated in Grades 2and 4. Three basicquestions were stated:

Ql: Do significantlymore second and fourth graderscomprehend material written withfrequent oral languagepatterns better than material writtenwith infrequent orallanguage patterns ? Q 2: Do fourth graderscomprehend materialwritten with frequent and infrequent orallanguage patterns significantlybetter than second graders ? Q3: Do girls comprehendmaterial written withfrequent and infrequent oral languagepatterns significantly betterthan boys? Results relevant to thenine hypothesesare grouped according to the three questions. The instruments usedto measure readingcomprehension of select oral languagepatterns were devised bythe investigator forthe present study. Because thetests were revised afterpreliminary piloting, additional estimates ofreliability and validitywere obtained after final data collection. Reliabilityand validityare discussed in the presentchapter so that the results of thestudy can be consideredalong with this information about the measuringinstruments.

69 RESULTS OF THE TESTS OF HYPOTHESES

Results Related to Reading Comprehension of Select Oral Language Patterns In this section the results relating to Question 1, whichconcerned the differential effect of frequent and infrequent oral language patterns on reading comprehension, are presented. HI: The number of second and fourth graders whose reading comprehension of material written with frequent oral language patterns is better than reading comprehension of material written with infrequent oral language patterns is not significantly greater than the number of second and fourth graders whose reading comprehension of material written with frequent oral language patterns is not better than reading comprehension of material written with infrequent oral language patterns. For the first chi square analysis, scores on Test A (frequent oral language patterns) and Test B (infrequent oral language patterns) were compared for all 300 subjects to see whether the number of higher scores on the former was significant. The results of the chi squareanalysis, with 1 degree of freedom anda =.0 5, arepresented in Table 14. As shown in the table, more subjects attained a higher reading comprehension score on Test A than on Test B (192 subjects who did, compared to 108;who did not). The number was significant at the .001 level; the null hypothesis was rejected. See Appendix D for the subjects' scores on Test A and Test B. For further clarification of this finding, Table 15 was constructed to show the mean score, range, variance, and standard deviation for all the subjects on Test A and Test B. As indicated, the mean score on Test A was higher than on Test B (31.37 for Test A and 29.84 for Test B). The range of scores on both tests was the same and the variance was similar,

70 Table 15 Mean, Range, Variance and Standard Deviation for All 300 Subjects on Test A and Test B

Readinc Comprehension Standard Test Mean Range Variance Deviation Test A (frequent oral language patterns 31.37 10-37 18.90 4.35 Test B (infrequent oral language patterns) 29.84 10-37 20.94 4.58 0......

although scoreson Test B varied slightly more than scores on Test A (the variance was 20.94 on Test B and 18.90on Test A). Frequency distributions of total scores on Test A and Test B forall subjects are presented in Appendix E. Because it was felt that considering the totalgroup might have con- cealed different results withinsome of?, smaller groups of interest in the present study, chi square analyseswere performed separately on test scores of subjects who were grouped bysex and by grade. The six additional analyses involved treating (1) each gradeas a separate group, (2) each sex within grade as a separategroup, and (3) each sex across grades asa separate group. The respective findingsare presented below. Each grade as a separategr Dt12. The result of the chi square analysis for the second graders is summarized in Table14, which shows that the chi square value of 15.86 was significant at the .01 level.More subjects obtained their higherscore on Test A than Test B (107 subjects' who did; 56 who did not).

72 The result of the chi square analysis for the fourth grade is presented in Table 14,its shown, the greater number of subjects obtained their higher score on Test A than Test B (85 subjects who did; 52 who did not). --The = chi = square-value was -significant at the .01 level. Each sex as a separate group within Grades 2 and 4. Within thesec- ond grade, test scores were treated separately for girls and boys. The findings for the former are presented in Table 14. The chisquare value of 6.47, with 1 degree of freedom, was significant at the .01 level. More subjects obtained their higherscore on Test A than Test B (5 2 subjects who did; 29 who did not). In Table 14, the results of the chi square analysisare presented for second grade boys. The table shows that the chisquare value of 9.5 6, with 1 degree of freedom, was significant at the .01 level.A greater number of subjects scored higher on Test A than Test B (55 subjects who scored higheron'T-est A; :27 who did not). Test scores were also analyzed separately for girls andboys within the fourth grade. -Table 14 summarizes the findings for thefourth grade girls. The chi square value of 7.57, with 1 degreeof freedom, reached the .01 level of significance. More subjects obtained theirhigher score on Test A than Test B (46 subjects who did; 23 who did not). The result of the chi square analysis for fourth grade boysis presented in Table 14. Aq shown in the table,more boys obtained their higher score on Test A than Test B (39 subjects who did;29 who did not), but the chi square value of 1.47, with 1 degree of freedom,was not significant at the .05 level.

Sex across Grades 2 and 4. The second- and fourth-grade girlsand the second- and fourth-grade boys formed twogroups which were treated separately. Table 14 summarizes the results of thechi square analysis

73 of the girls' scores. As shown in the table, more girls obtained their higher score on Test A than Test B (98 subjects who did; 52 who did not). The chi square value of 15.36, with 1 degree of freedom, was significant at the .001 level.- More of the second- and fourth-grade boys obtained their higher score on Test A than Test B (94 subjects who did; 56 subjects who did se, not). As indicated in Table 14, the chi square value of 9.62, with 1 degree of freedom, was -significant at the .01 level.

Results Related to Grade Level In this section, results are presented for the two hypotheses that were derived from Question 2, which focused on the relationship between grade level and reading comprehension of material written with frequent and infrequent oral language patterns. In eachcase, the difference between the mean reading comprehension test scores of the twogroups being compared was tested for significance; the t testwas used for this purpose. In accordance with the requirements of the null hypothesis,the two-tailed t test was used. The .05 level of significancewas accepted. A null hypothesis was rejected, consequently, if t > t 1 - (a/2)[N1 + N2-2] or if t < ta/2[N1 + N2-2]

Because the variance of scores on both reading comprehensiontests differed from grade to grade, F ratioswere computed to test the homogeneity of variance. The F ratio was significant at the .05 level for grade level comparisons on each reading comprehensiontest; the model for the t test which is appropriate for unequal variancewas used. H 2: Fourth graders do not do significantly better than second graders in their reading comprehension of materialwritten with frequent oral language patterns.

74 The results of the 't test for secondand fourth gradersare presented in Table 16 for Test A (frequent orallanguage patterns)As shown in the table, the difference betweenthe second- and fourth-grademeans on Test A was significant at the .01level (a mean of 32.98 forGrade 4 and 30.02 for Grade 2). The nullhypothesis was consequentlyrejected. Frequency distributions of totalscores on Test A are presented inAppendix E for both grades.

H3: Fourth graders donot do significantly better thansecond graders in their readingcomprehension of material written with infrequent oral languagepatterns.

Table 16 summarizes theresults of the t test forsecond and fourth graders on Test B (infrequentoral language patterns).The difference in the second- and fourth-grademeans was significant at the .01 level (a mean of 31.63 for Grade 4 and28.34 for Grade 2). The nullhypothesis was rejected. In Appendix E,frequency distributions oftotal test scores on Test B are presented for each grade. Table 17 was constructedto elaborate the findingspresented for Hypotheses 2 and 3.In the table, themean, range, variance, and standard deviation ofscores on both tests are comparedfor the two grades.It is clear that in additionto the lower meanscore, the second graders had a greater range, variance, andstandard deviationon Test A and Test B.

Results Related to Sex

In this section, resultsare discussed for the six hypothesesthat were derived from Question 3,which dealt with therelationship ofsex and reading comprehensionof material written withfrequent and infrequent oral language patterns.'Fot each comparison ofthe meanscores, the t test models thatassume either equal or unequalpopulation variancewere

75 Results of the t test of Mean Scores for Second and Fourth Graders Table 16 TestReadingComprehension N SecondMean Grade SD N FourthMean Grade SD Differ-Mean, ence value t- Signifi-canceLevel Hypothesis i languageoralTestpatterns)(frequent A 163 30.02 4.96 137 32.98 2.72 2.96 6.58 <.01 H2 rejected languageoralTestpatterns)(infrequent B 163 28.34 5.06 137 31.63 3,10 2.29 4.87 <.01 H3 rejected Table 17 A Comparison of the Mean, Range, Variance and Standard Deviation of Scores for Grades 2 and 4

Reading Comprehension Grade Standard Test Level Mean Range Variance Deviation

Test A (frequent oral 2 30.02-10737 24.65 4.96 language patterns) 4 32.98 19-37 7.39 2.72 Test B (infrequent oral 2 28.34 11-36 25.62 5.06 language patterns) 4 31.63 16-37 9.59 3.10 used depending on whether or not the computed F ratio revealed significant differences in the variances of boys and girls. In accordance with the requirement of the null hypothesis the two- tailed.t test was used. The .05 level of significance was accepted. A null hypothesis-was rejected if t > t [1\1 + N-2] or if t < ta/2 [N]. + N2 - 2]. 1 (a/2) 1 2

H4: Second and fourth !grade girls do not do significantly better than second and fourth grade boys in their reading comprehension of material written with frequent oral language patterns. The variances for the girls and boys differed significantly at the .05 level according to the F ratio. Thus the t test model, appropriate for unequal variance was used. Results of the t test are presented in Table I 8. As shown in the table, the difference between the mean scores of the girls and boys on Test A (frequent oral language patterns) was significant at the .02 level. The higher mean favored the girls (31.97 for the girls and 30.77 for the boys). The null hypothesis was consequently rejected.

77 Results of the t Test of Mean Scoies for Boys Table 18 and Girls Across Grade Levels TestReadingComprehension N Mean Girls SD N Mean Boys SD Differ-enceMean t- value Signifi-Levelcance Hypothesis orallanguageTestpatterns)(frequent A 150 31.97 3.77 150 30.77 4.79 1.21 2.46 <.02 H4 rejected oralTestpatterns)language,(infrequent B 150 30.27 4.29 150 29.41 4.82 .85 1.61 H5 accepted H5: Second- and fourth-grade girls do not do significantly better than second- and fourth-grade boys in their reading comprehension of material written with infrequent oral language patterns.

According to the computed F ratio,the variances of the girls and boys did not differ significantly at the .05 level. The t test model which assumes equal variances was used. As shown in Table 18, the difference between the mean scores of girls and boys on Test B (infrequent oral language patterns) was not significant at the .05 level, but the higher mean was obtained by the girls (30.27 for the girls and 29.41 for the boys). The- null hypothesis was accepted. H6: Second-grade girls do not do significantly better than second-grade boys in their reading comprehension of material written with frequent oral language patterns. To test this hypothesis, the t test model which assumes unequal variance was used because the F ratio was significant at the .05 level. The mean scores for Test A (frequent oral language patterns) indicated that second grade girls scored higher than second grade boys (30.46 for the girls and 29.59 for the boys). As shown in Table 19, however, this difference was not significant at the .05 level. The null hypothesis was accepted. H7: Second-grade girls do not do significantly better than second-grade boys in their reading comprehension of material written with infrequent oral language patterns. Since the F ratio of the two variances was not significant at the .05 level, the t test model which assumes equal variances was used. Results of the t test for second grade boys and girls on Test B (infrequent oral language patterns) are presented in Table 19. Although the higher mean was attained by the girls (28.75 for the girls and 27.93 for the boys),the difference was not significant at the .05 level. The null hypothesis was accepted.

79 Results of the t Test of Mean Scores for Boys and Girls in-Secdnd Grade Table 19 TestReadingComprehension N Mean Girls SD N Mean Boys SD Differ-enceMean t value' 5 Levelcance r Hypothesis oralpatterns)languageTest(frequent A 81 30.46 4.23 82 29.59 5.59 .87 1.15 H6 accepted languageoralTestpatterns)(infrequent B 81 28.75 4.74. 82 27.93 5.35 .83 1.03 H7 accepted H8: Fourth-grade girls do not do significantly betterthan fourth-grade boys in their readingcomprehension of material written with frequent oral languagepatterns. The F ratio of the two varianceswas significant at the .05 level; the t test model whichassumes unequal variances was used. As shown in Table 20, the mean scores for fourth-gradegirls and boys were not significantly different at the .05 level.However, the higher mean score was obtained by the girls (33.75 for the girlsand 32.19 for the boys). The null hypothesis was accepted.

H9: Fourth-grade girls do not do significantlybetter than fourth-grade boys in their readingcomprehension of material written with infrequent orallanguage patterns. Because the F ratio of the varianceswas significant at the .05 level, the t test model whichassumes unequal variance was used. Results of the t test for fourth-grade girlsand boys on Test B (infrequent orallanguage patterns) are summarized in Table 20.The table indicates that although the highermean score was attained by the fourth grade girls (32.04 for the girls and 31.21for the boys),the difference was not significant at the .05 level. The nullhypothesis was accepted. To elaborate the findings forHypotheses 4-9, which dealt with the relation of sex to readingcomprehension of select oral languagepatterns, Table 21 was constructed.In the table, themean, range, variance,and standard deviation for Test A andTest B are summarized for girls andboys within and across grade levels.From inspection of the table it is clear that in addition to obtainingconsistently lowermean scores than the girls, the boys were characterized bygreater variance and a greaterrange of scores. Two exceptions to the latterwere the total group of boys and the second-grade boys who, in Test B, hada smaller range of scores than the corresponding group of girls.

81 Results of the, t Test of Mean Scores for Boys, and Girls im Fourth Grade Table 20 TestReadingComprehension N Mean Girls SD N Mean Boys SD Differ-Meanence valuet- Signifi-Levelcance Hypothesis oralpatterns)languageTest(frequent A 69 33.75 2.04 68 32.19 "3.09 .56 1.25 H8 accepted languageoralTestpatterns)(infrequent B 69 32.04 2.81 68 31.21 3.33 .84 1.60 H9 accepted Table 21 Reading Mean, Range, Variance, and Standard Deviation of Scores on Test A and Test B for Girls and Boys Within and Across Grades Two and Four TestTestComprehension(frequent A AllSubjects boysgirls 150N Mean30.7731.97 . Range1014-37 -3.7 Variance 14.2022.99 DeviationStandard 4.793.77 oralpatterns) language Fourth-gradeSecond-grade girls boysgirli 828169 33.7530.4629.59 10-3714-3628-37 31.2817.85 4.16 5.594.232.04 oralTest (infrequentlanguageB AllFourth boysgirls -grade boys 150 68 30.2732.1929.41 13-3611-3721-37 18.4023.25 9.58 4.824.293.09 patttrns) Fourth-gradeSecond-grade boysgirls boysgirls .698281 31.2132.0427.9328.75 17-3613-3411-3721-37 11.0928.6622.51 7.90 5.353.332.814.74 Summary of Results The results for each of the nine hypotheses are summarized as follows:

Hypothesis 1.In general, significantly more second and fourth graders obtained higher reading comprehension scores on material written with frequent oral language patterns (Test A) than on material written with infrequent oral language patterns (Test B). The exception was fourth- grade boys, whose higher scores on Test A did not reach an acceptable level of significance. Hypothesis 2.Fourth graders comprehended material written with frequent oral language patterns (Test A) significantly better than second graders. Hypothesis 3.Fourth graders comprehended material written with infrequent oral language patterns (Test B) significantly better than second graders. Hypothesis 4. Second- and fourth-grade girls comprehended material written with frequent oral language patterns (Test A) significantly better than second- and fourth-grade boys. Hypothesis 5. Second- and fourth-grade girls did not do significantly better than second- and fourth-grade boys in their reading comprehension of material written with infrequent oral language patterns (Test B). Hypothesis 6. Second-grade girls did not do significantly better than second -grade boys in their reading comprehension of material written with frequent oral language patterns (Test A). Hypothesis 7.Second-grade girls did not do significantly better than second-grade boys in their reading comprehension of material written with infrequent oral language patterns (Test B)

84 Hypottre fourth-grade boys in their reading comprehensionof material written with frequent oral language patterns (Test A).

Hypothesis 9. Fourth-grade girls didnot do significantly better than fourth-grade boys in their readingcomprehension of material written with infrequent oral language patterns (TestB).

A summary of the results of thetests of hypotheses is presented in Table 22.

DISCUSSION OF THERESULTS

A discussion of the results followsfor the hypothesis relatedto reading comprehension of selectoral language patterns,the hypotheses related to grade level, andthe hypotheses relatedto sex.

Results Related to ReadingComprehension of Select Oral LanguagePatterns Rejection of the first hypothesiswas expected on the basis of RuddelPs research (1963).With the results of Hypothesis1, Ruddell's major finding was confirmed,elaborated, and extendedin three ways: (1) applicability- ofthe finding .to childrenat the primary as wellas the intermediate level of elementaryschool was demonstrated; (2)the finding was based'on measuring instrumentswhich did hot systematicallybias the perforMance of childrenon the infrequent oral language patterns;1(3) the

1 As previously discussed in. thereview of RuddelPs study (1963)in Chapter I, it wo.o suggestedthat his use of the clozeprocedure may have biased his results.

85 Table 22 Statistic Summary of the Results of the Tests of Hypotheses df Signifi-Levelcance Hypoth- x2Used GroupAll subjects Considered 11 <.001 esis Hi Decision reject SecondSecond-gradeFourth gradersgraders boysgirls 1 <.<.01<.05 01 AllFourth-grade boysgirls boysgirls i1 <.01<.001MI MI MI ON t test Second and fourth graders (Test B)A) 29298 8 < 0 1 * 0 1 H3H2 reject"reject' All girls and all boys (Test A) 29298 8

ill = average readers; 7-9 (corresponding to the 6.0-9.5 grade placement scores) = good readers. The obtained frequencies in each category are presented in Table 23 for girls and boys. As indicated, the scores for both sexes were similarly rated; there is no substantial evidence to support the second explanation. Clearly more research is needed to clarify the performance of fourth-grade boys in the present study.

Table 23 The Number of Poor, Average, and Good Readers Among Fourth-Grade Girls and Boys as Determined by Scores from the Paragraph Meaning Section of the Stanford Achievement Test

Poor Readers Average Readers Good Readers (Stanine 1-3) (Stanine 4-6) (Stanine 7-9)

Girls 7 38 24 Boys 35 27

The Results Related to Grade Level Significant differences between the performance.of second and fourth graders on both reading comprehension tests were expected. Possibleex- planations for these differences are listed and discussed below. 1. Fourth graders read better than second graders. This is -not a profound nor contested observation about readingper- formance of children in successively higher grade levels; it is basedon the fact that reading is a skill that improves with practice and experience. An interesting problem is raised by the tests used in the present study, however, because they were not primarily intended tomeasure skills that are taught in reading programs at the elementary school. Typical measures of reading ability such as vocabulary and reading speedwere carefully controlled in the materials devised for the studyso that they would not be heavily weighted factors in the children's reading comprehensionscores. As discussed more fully in Chapter II, these additional controlswere imposed to limit the effect of distinctly different levels of readingachieve- ment between the two grades prior to testing:(1) no time limit so that second graders would not be penalized for taking longerto read through the tests;(2) practice items to equalize thesubjects' experience with the particular nature of the reading comprehension task; (3)standardized, clear directions for all subjects; (4)a reasonable test length which, in piloting, was shown not to overtax the endurancelimitations of the younger subjects; (5) the use of words whichare presumed relatively easy for primary grade children (predominantly from the Stone List); (6) directionsfor children to ask for help withany word they did not know; (7) the use of a measure of reading comprehension thatwas assumed to be relatively new and interesting to subjects of both grades. If it can be assumed that differencesbetween the grades in terms of decoding ability were minimized rather thanmaximized by the controls imposed on the materials, then the firstexplanationfourth graders read better than second gradersis rejectedas it is presently worded and a second explanation is proposedas a more acceptable alternative: 2. Second graders are less successfulthan fourth graders in com- prehending the relationship between oraland written language patterns. From the greater variability of second-gradescores on both Test A and Test B, it is apparent that childrenat this level of reading ability and experience have not yet developedas much flexibility in handling diverse language patterns in written materialas fourth graders. Reasons for this are beyond the scope of the present study, butsome suggestions are possible from two areas:(1) a linguistic analysis of the;Lading process, and (2) researchon children's written language structures.

89 With respect to the first area,Lefevre (1964) has emphasized the need to read for intonation patterns in sentencesto comprehend the "melodies of the printed page." From the very beginning of reading instruction, therefore, a child must learn to read sentences; beginning with isolated words results in the unnatural practice ofgiving a full stress to each wordthe opposite of what is done in speaking. An early understanding of the relation between spoken andwritten language is consequently hampered. From this point of view, thedifferent performance of second and fourth graders might have been due, in part, to the still incomplete understanding by the former of (1) therelation between spoken AO written language, and (2) the constraints ofwritten language that need to be understood and compensated for,examples being spatial rather than temporal orientation of language,and missing intonation clues. From the area of researchon children's written language structures, Hunt (1965) has concluded that fourth graders understand complexgram- matical structures because they write themoften and accurately. The fact that these structureswere not produced as frequently or with as-much variety by fourth graders as by the older childrenin his study, however, led Hunt to recommend thata program should be developed to Widen the younger children's span of grammatical attention andconcern. The intent would be to work toward performancethat is comparable to that of twelfth graders. The implied difference in grammaticalawareness betWeen the grades in Hunt's study is applicable to thedistinction that 'can be made between , the grades studied in this investigation.Compared to fourth graders, second graders might not beas consciously aware of how word order in 42. the visual context of reading relatesto meaning even though theyare capable of producinga variety of complex word orders in their oral language.

90 Indeed, second graders might beso busy looking at and considering each word that they miss the meaning of wholelanguage patterns.

3. The materials were too easy for fourth graders. This explanation could be possible forone of two reasons:(1) either the materials were too easy because theywere written at the second-grade level, or (2) the materialswere too easy because the patterns themselves were too easy for children at the fourth-grade level. Withrespect to the latter, it is suggested that the testswere not as reliable measures of reading comprehension for fourth gradersbecause they tested something, defined as language pattern difficulty,which was not as relevanta concern for this grade as for second grade. 2 Perhaps theinfrequent oral language patterns selected for Test Bwere not infrequent enough for the older subjects; selection of less frequentpatterns might have increased the difficulty of the items in TestB. Acceptance of either of the two alternativessuggested above, however, does not affect the fact that despite the relativeease with which fourth graders handled the materials,a significantly greater number of them obtained higherscores on Test A than on Test B. The implications of the results of the first hypothesis,which stated that the number of second and fourth graders whoobtained higher scoreson Test A than Test B would not be significant, shouldnot be restricted to second graders.

Results Related to Sex

The general finding that boys didless well on each test than girls was expected on the basis ofmany researchers who have found differential

2 See the discussion of test reliabilityin this chapter and the results of item analysis by grade level,found in Appendix F, for furtherclarifica- tion of this possibility.

91 performance of the sexes in elementary school reading andlanguage related tasks.This difference due to sex has been noted from thevery beginning stages of reading readiness, where girls havegenerally been found -superior (Barrett, 1962).Broader studies of the relation ofsex to performance in several language tasks have generallyfound differences, though not all of them have been consistentor significant ones.3 Loban's (1963) finding that boys who performed poorlyin language obtained the lowest readingscores while boys who did well in language obtained among the highest readingscores was supported, with few exceptions,by the results of the six hypotheses relatedto sex in the present study. The poorest scores for all 300 subjectswas attained by a boy (10, on Test A), and the highestpossible score on each test, 37,was also attained by a boy. These resultsemphasize the need to consider great variation of ability within eachsex rather than treating members of one sex as a homogeneous group.

Reliability of the Reading ComprehensionTests A broad definition of test reliability includesreference to these three basic characteristics ofa test (Cronbach, 1960):(1) the extent to which the measuring instrument produces similarresults after repeateduse (stability and dependability); (2) theextent to which errors ofmeasure- ment are present (random chance or error); and (3)the extent to which the given results are a truemeasure of what is being assessed (accuracy). In

3 McCarthy (1954) found girls superiorto boys in most aspects of language, but Templin (1957), Strickland(1962),Loban (1963), and Riling (1965) did not find clear-cut,consistent sex differences.

92 accordance with this definition, three differentmeasures of reliability were obtained for each reading comprehension test; the results of each are discussed below. ---Firstrreliability was examined- in- terms of stability. Dueto the limited scope of the study and restrictions of time, the usualtest-retest method was not feasible. Instead, stabilityacross rather than within tests was evaluated by correlatingscores on Test A and Test B to see to What extent good or poor performanceon one was indicative of good or poor' performanceon the other. For this purpose, the Pearson product- Moment coefficient of correlation was determined. This coefficientmeas- ured the- amount of relationship between eachset of scores on Test A and Test B; itwas based on the products of pairs of these scores .4 The obtained coefficient of correlationwas .83, which was significant at the .01 level. Therefore, performanceon Test A was significantly related to performance on Test B. Second, reliability was assessed interms of the standard error of measurement, which provides an indication of the reliabilityof individual scores (Lindvall, 1967). Unlike the reliability coefficient, the standard error of a score is independent of the variability of thegroup and consequently is not affected bygroup homogeneity, a characteristic that appre- ciably lowers the reliability coefficient (Blooxn,1966). The standarderrors for Test A and Test Bare summarized in Table 24 both within andacross

4Thebasic' formula for the Pearson product-momentcorrelation is Exy 111=1 where x = (X- X), y= - Y) , and x and 2 2 (Ex) (Zy) y are the two variables being considered (Tate, 1955,p. 238).

93 Table 24 Hoyt Reliability Coefficients Within and Across Grades for Test A and Test B Grad Level Second & Test Second Fourth Fourth

R SE R SE R SE

:Test A (frequentoral language patterns) .81 2.14 .58 1.74 .79 1.97

Test B (infrequent oral language patterns) .81 2.20 .59 1.95 .78 2.10

Grades 2 and 4.Inspection-of the table reveals that (1) the standard errors for Test B within and across grades were somewhat higher than for Test A, and (2) the standard errors for the second gradewere somewhat higher on each test than those for the fourth grade. Since themaximum standard error was 2.20, the-testswere considered fairly reliable measures of individual scores. Third, the degree of internalconsistency, or accuracy, was determined by using the Hoyt Reliability Coefficient.5 The reliabilitycoef- ficients for each test within and across Grades 2 and 4are summarized in Table 24. On the basis of information presented in the table,these .1141.11

5 The Hoyt Reliability Coefficientwas obtained from a 1604 computer program, FoRTAP. (Baker and Martin, 1968).

94 observations' can be made:(1) Within each grade, reliability coefficients of Test A and Test B were similar;(2) the reliability coefficients of both tests were considerably higher for second-gradesubjects than fourth-grade subjects. The lOWtest reliability for the fouithgrade6-can be examined in light of the findings for the tests of hypotheses, discussed in the preceding section of this 6hapter:, rrom the resultsrelated-,to grade level differefices in teStiSeriOrniiiiCe 012 and 1.11):, it-is obvious' in Table 17 that the vari- . ance of ,scores for the second grade was almost three times greater on Test A thafefoi fourth-grade and more =than three times greater for the second grade on Test B. Since a reliability coefficient depends on the variability of scores (Anastasi, 1960),-lack of variability at the fourth grade level probably contributed heavily to the low reliability coefficient of each test for fourth grade subjects. From the means and ranges of scores for, the fourth grade , information also presented in Table 17, it was apparent that scores were not distributed at both ends of the scale but were, instead, clustered near the top (see Appendix F for the frequency distribution of scores for the fourth grade). Tests as negatively skewed as these for fourth graders may have been satisfactory for measuring differences at the lower end of the group of subjects, but not for measuring differences at the upper end; the test could not -distinguish between scores of subjects who tied at or near the perfect score of 37. Other factors that can affect test reliability such as the testing environment, instructions, and time limit (Anastasi, 1961) were controlled

6Bloom (1966) has stated that a reliability coefficient of .60 is adequate for group testing.

95 across grades in that (1) allsubjects were tested within theirown, classroom, a familiar environment, (2) thc: instructions were standardized, and (3) there was no time limit. These factors probably did. not contribute to -the diSarepanCY in "reliability froth 'grade to grade. In conclusion, the considerably smaller variance of scores obtained on both Test A and Test B by fourth graders probably accounted in large part for the low reliability coefficients at the fourth grade level. Explana- tions for the smaller variance and possible implications were presented in the discussion of the results that preceded this section of the chapter. See Appendix F for results of an item analysis of Test A and Test B An terms of item reliability (R) and difficulty (P) for each grade.

Summary Results of three measures of reliability were discussed for Test A and Test B. Qn the basis of these results it seemed that the tests were fairly reliable measures of individual scores. The tests were acceptedas reliable instruments for the second-grade subjects but considerably less reliable instruments for the fourth-grade subjects. The smaller variance ofscores for the latter probably contributed substantially to the lower reliability.

Validity of the Reading Comprehension Tests Two types of validity were considered in deVising and evaluating the two reading comprehension tests. They were construct validity, which has to do with the meaning of test scores in terms of psychological concepts and theory (Cronbach, 1960), and content validity, which has to do with the appropriateness of the content of the test items in terms of what the test as a whole is measuring (Lindvall, 1967).In this section each of the two kinds of validity is discussed.

96 Construct Validity

Theories of languagelearning and its relationto reading achievement have not yet been tested thoroughly.Researchers have madeinteresting but inconclusive reference to the factthat children whoare good readers have considerable facility with spokenlanguage; thesechildren usemore complex language structures and theyuse the basic structureswith flexibility (Strickland, 1962; Loban, 1963).Conversely, childrenwho are poor readersseem to lack the kind of-flexibility withpatterns of languagestruc- ture that characterizesmature language. Learning theorists and educators generallysupport the idea ofmoving from the familiar to the unfamiliar inteaching children.7With respect to reading, a task which' involves learningto obtain meaningfrom written language 'patterns, it is also logicalto move from thefamiliar to theun- familiar. In a sense; this is alreadydone. Rigid controlover vocabulary is typically imposed in basalreaders; a littlevocabulary is developed over a-long period oftime.In terms of thecontext of thesewordsthe sentence or language patternhowever, suchcontrol is notapparent (Strickland,1962; Riling, 1965). It is reasonable to suggest that learningthe relation betweenwritten 'and spoken language would be facilitatedgreatly if childrenbegan reading language patterns that were familiarto them beforethey moved, gradually, to less familiar patterns. Recentresearch on children'soral language

7 Lindgren (1962)provides a typical cussion of the example of this ideain his dis- conceptual approachto learning:".. forced when the .learning isrein- various experiencesof life. are seen by the students as interrelated. Informationand skillsare more readily learned tained.ifwe see how they"are and re- related to otherthings we know" (p.267).

97 provides a reasonable way to decide whetheror not certain patterns are familiar or unfamiliar to children: decisionscan be based on observations of oral language patterns that appear very frequentlyor very infrequently in their speech. Thus, 1.;-, is suggested that theterms "familiar" and "unfamiliar be made synonymous, respectively, with"frequent" and "infrequent" patterns as they appear in children's oral language. In this manner, then, the theoretical framework for thepresent study was derived from two sources:(1) learning theory which emphasizesmov- ing from the familiar to the unfamiliar formore effective and efficient learning, and (2) the assumption thatone can determine language patterns for written material which are familiaror unfamiliar to children by observ- ing which patterns childrenuse frequently or infrequently in their oral language. From this framework,a major hypothesis of the study was stated: Hl:The number of second and fourth graders whosereading comprehension of material written with frequent oral language patterns is better than reading comprehension of material written with infrequent oral languaaepat- terns is not significantly greater than the number of second and fourth graders whose readingcomprehension of material written with frequent oral language patterns is not better than reading comprehension of material written with infrequent oral languagepatterns. Support for the reason to suspect differencesin reading comprehension of frequent (familiar) and infrequent (unfamiliar)patterns was derived from basically twosources. First, from an observation of research in this area (Ruddell, 1963; Tatham, 1967; Tatham, 1968), itwas clear that there is unanimous though limited evidence that childrencomprehend material written with frequent oral language patterns significantlybetter than material written with infrequent oral languagepatterns. 8 Second, from their

8 Fourth graders were the subjects for each ofthe three studies.

98 analysis of children's oral language, some researchers have indicated that certain patterns are the building blocks of children's oral language; they are used frequently regardless of age or intelligence.Since these patterns are common, one would expect children to have the least difficulty reading them. On the other hand, since the use of infrequent patterns is not common to all children, one would expect that the ability to read them would distinguish between levels of achievement more than the ability to read the "easier" frequent patterns. A logical hypothesis that emerges from this line of reasoning is that scores on Test B (infrequent patterns) should be a better predictor of general reading comprehension, as measured by the Stanford subtest, than scores on Test A (frequent patterns). To determine whether or not this was the case, a multiple regression analysis was performed by means of REGANI.9 For the analysis, the two independent variables were scores on Test A and Test B; the dependent variable was scores from the Stanford subtest. As shown in Table 25, the partial regression coefficients of both Test A and Test B were significant at the .01 level.In other words, both independent variables contributed significantly to the scores on the Stanford subtest. To look at this another way, partial correlations of the Stanford subtest with each of the reading comprehension tests were determined as part of the REGANI program. The partial correlations were quite low, as indicated in Table 26. Thismeans that Test A and Test B apparently had a great deal incommon. From its higher partial correlation with the Stanford subtest, however, Test Bpro- vided evidence of an element that was unique to it alone. This unique element, probably difficulty, correlated more highly with the Stanford subtest than Test A.

9Susanta Guha (1966). Given N number of observationson p variables, REGANI can be used to solve linear regression equations by the method of least squares.

99 Table 25 Partial Regression Coefficients of Test A and Test B

Partial Regression Variable "COefficietitS t-value

Stanford subtest -32.76 6 19 -5.29 Test A 1.01 .35 2.88** Test B 1.44 .34 4.26***

* < .01 *** < .001

Table 26 Partial Correlations of Test A and Test B with the Paragraph Meaning Section of the Stanford Achievement Test

Test Partial Correlation

Test A .17** Test B .25**

.01

In conclusion, from both the partial correlations and the multiple regression analysis it would seem that performance on Test B rather than Test A was a somewhat better indicator of general reading comprehension as measured by the Stanford subtest. A last statistical measure of validity was obtained by determining the coefficient of correlation between Test A, Test B, the combined scores, and the Stanford subtest. The resulting correlation matrix is summarized in Table 27. As noted in the table, all the correlations were significant at the .01 level. 100 Table 27 Validity Coefficient of Correlation between Test A, Test B, the Combined Scores and the Paragraph Meaning Section of the Stanford Achievement Test

1 2 3 4

Test A 1.00 .83** .95** .56** Test B 1.00 .95** .58** Combined Scores (A I-, B) 1.00 .61** Stanford Subtest 1.00

** p<01 "ir

The high intercorrelation of Test A with Test B (.83) could be inter- preted in at least one of three ways (Cronbach, 1960):(1) reading corn- prehension of frequent oral language patterns (Test A) caused or influenced reading comprehension of infrequent patterns (Test B);(2) reading comprehension of infrequent oral language patterns (Test B) caused or influenced reading comprehension of frequent patterns (Test A); or (3) scores on both Test A and Test B were influenced by some common factor or factors. From inspection of the coefficients of correlation of Test A and Test B with the Stanford subtest, it is evident that general reading ability, defined by scores on the Stanford subtest, was a factor in determining performanpe on the materials written with frequent and infrequent oral language patterns .

Content Validity As discussed in Chapter II, the frequent and infrequent oral language patterns were selected from among those that appeared in the oral language of both second and fourth graders, according to the findings of Strickland's

101 study (1962).Since Strickland reporteda total of over 200 different patterns, it was obvious that onlya select number could be used in the kind of tests devised for the present study. Two small samples from the totaluniverse of children's oral language patterns were designatedas "frequent" and "infrequent," and from within each sample, a smaller numberof patterns was selected foruse in the study.It was assumed that findingsfrom the subsamples could begen- eralized to the samples and,finally, to the universe ofpatterns labeled "frequent" and "infrequent"in the oral language of children.Clearly, then, the patterns selected foreach test were representative butby no means inclusive. Within a language pattern therewere several ways to fill in each slot or movable. As previouslystated, a decisionwas made to include a variety of fillers suchas adjectives, phrases, clauses, andcompound nouns, verbs, or objects. Once again, thetypes of subordination and grammatical complexities thatwere used were representative butnot inclusive of the possibleways of filling in each slot and movable. As a final checkon content validity, a judge whowas thoroughly familiar with Strickland'swork and similar studies oforal language pat- , terns 'analyzed the 37 itemsin each reading comprehensiontest to be sure that the patterns used correspondedwith the patterns that hadbeen selected for use by theinvestigator.

Summary

Construct and content validitywere discussed for the two measuring instruments devised for thestudy by the investigator.Both separately and together, thetests seemed to be basedon valid, logical assumptions from learning theory andrecent studies of children's orallanguage and

102 its relationship to reading. In terms of content, the tests were representative of patterns designated "frequent" and "infrequent" in the oral language of second and fourth graders.

Summary

Results of the tests of the nine hypotheses of the studywere presented and discussed in this chapter. In addition, obtained estimates of the reliability and validity of the two measuring instruments devised by the investigator were presented.

...

103 IV SUMMARY,CONCLUSIONS, AND IMPLICATIONS

Background and Purpose The relationship between reading comprehension and material written with frequent and infrequent oral language patterns of children was investigated in the present study. Linguists have emphasized that the area of reading instruction must be concerned much more than it is with understanding language and how it is used in spoken and written messages (Carroll, 1964)that is,how spoken and written language are related. For many children the process of learning this concept is made unnecessarily difficult by the "unnatural language". of their reading materials.Vocabulary is typically rigidly controlled in their texts, but there is no apparent evidence of control over the context of wordsthe language patterns which the children read (Strickland, 1962; Riling, 1965) Writing materials for children which are structured more like the way they speak is a linguistically sound alternative to obstructing children's understanding of the fundamenItal relationship between the language they bring to school and thelangualgethey encounter in written material.In the single study which investigated this possibility, fourth graders comprehended material written with patterns children use frequently in their oral language significantly better than material written with patterns children use infrequently in their oral language (Ruddell, 1963).More research is needed to confirm this finding for different groups of children

104 at different grade levels. Determining whetheror not certain oral language patterns used by children are easy or difficult for them to comprehendin written material could bring a relevant but heretofore neglectedcharacter- istic of the readerhis oral languageto the concept ofreadability. The present study was undertaken to investigate furtherthe relationship between reading comprehension and materialwritten with select oral language patterns at Grades two and four. Therelation of sex to reading comprehension of these patternswas of additional interest. More specifically, the study was designed toanswer three basic questions:

Do significantly more second and fourth graderscomprehend _material written with frequentoral language patterns better than material written with infrequent orallanguage patterns? Do fourth graders comprehend material writtenwith frequent and infrequent oral language patternssignificantly better than second crraders ? Q3: Do second and fourth grade girls comprehendmaterial, written with frequent and infrequent oral languagepatterns significantly better than second and fourth gradeboys ?

Related Research The review of the literaturewas divided into two sections: studies of elementary school children's orallanguage, and experimental studies using written material basedon an analysis of children's oral language. The first section was limited toinvestigations from 1959-1968 thatwere based on a broad analysis of oral languageobtained from relatively large samples of elementary school children.From the five studies conducted during this periOd, several conclusionswere drawn:(1) the researchers demonstrated that children's languagecan be analyzed by methods derived from structural linguisticsand transformational grammar and that information from these methodspermits a variety of language characteristics

105 to be described and compared for childrenOf-widely different abilities and age levels;(2) a relationship betweenreading comprehension and select oral language characteristics like theuse of subordination and movables Was siiggeite4 (3) from someof the findings, itwas apparent that chil- drenS Materialsare not written with patterns that childrenuse in their oral langUage. The second section of thereview dealt with the only studyin this area at present. Ruddell (1963) . constructed materials written withfrequent and infrequent oral languagepatterns of fourth graders,as determined by Strickland (1962), and he testedfourth graders' readingcomprehension of them. He found that childrencomprehended materials writtenwith the frequent patterns significantlybetter than materials thatused the infrequent patterns. Ruddell concludedthat reading comprehensionis a function of the similarity of oral andwritten language. Information from the five studiesof children's oral languageand the inquiry into children's readingcomprehension of select orallanguage patterns provided both the frameworkand the impetus for thepresent investigation.

Method

To determine whetheror not children comprehend materialwritten with frequent oral languagepatterns better than materialwritten with infrequent oral language. patterns,two reading comprehensiontests were devised by the investigatorTest A was written withpatterns that appear frequently in the oral language of bothsecond and fourth graders;Test B was written with language patterns thatappear infrequently in the oral languageof second and fourth graders. The patternswere selected from Strickland's (1962)tabulation of the frequency with which orallanguage patterns of secondand fourth graders

106 appear; five highly frequent patterns were selected for Test A and six less frequently used ones were selected for Test B.They appeared in proportion to the number of times childrenuse them in their oral language. Certain restrictions were imposedon the materials to prevent bias at either grade and to assure similar content andcomplexity (other than the variable of pattern complexity) fromtest to test. The variables of vocabulary, content, and grammatical complexitywere defined and then controlled within and across tests. Every effortwas made to make the actual decoding process as easyas possible for the second graders. Each test consisted of 37 items (sentences)which were selected after preliminary piloting according to (1) how wellthe distractors functioned, (2) how well the item correlatedwith the rest of the test, and (3) how dif- -ficult the itemwas .Three pictures were drawn for eachsentence; one illustrated the exact content of thesentence while the other two pictures differed in oneor more details.Reading comprehensionwas determined by the child's ability to read eachsentence and select ane of the three pictures that best represented thecontent of the sentence. Subjects for the study were drawn fromtwo elementary schools in Madison, Wisconsin. All classroomsof second and fourth graders in each school were used, yieldinga final total of 163 second graders, 81 girls and 82 boys, and 137 fourthgraders,69 girls and 68 boys. The subjects were tested by classroomfor three separate testingses- sions. During one week in the spring, thereading comprehension tests were administered on two consecutive morningsto each class. During the following week, the ParagraphMeaning section of the Stanford Achievement Test was administered.Subjects were randomly assigned to read either Test Aor Test B first, Derived from the three questions,nine hypotheses of the studywere tested:

107 Hl: The number of second and fourth graders whose reading comprehension of material written with frequent oral language patterns is better than reading comprehension of material written with infrequent oral language patterns is not significantly greater than the number of second-and fourth graders-vihog e reading Coinprehen- sion of material written with frequent oral language patterns is not better than reading comprehension of material written with infrequent oral language patterns. H 2: Fourth graders do not do significantly better than second graders in their reading comprehension of material written with frequent oral language Patterns. H3: Fourth graders do. not do significantly better than second graders in their reading comprehension of material written with infrequent oral language patterns. H4: Second and fourth grade girls do not do significantly better than second and fourth grade boys in their reading comprehension of material written with frequent oral language patterns. H5: Second and fourth grade girls do not do significantly better than second and fourth grade boys in their reading comprehension of material written with infrequent oral language patterns. H6: Second grade girls do not do significantly better than second grade boys in their reading comprehension of material written with frequent oral language patterns. H7: Second grade girls do not do significantly better than second grade boys in their reading comprehension of material written with infrequent oral language patterns. H8: Fourth grade girls do not do significantly better than fourth grade boys in their reading comprehension of material written with frequent oral language patterns. H9: Fourth grade girls do not do significantly better than fourth grade boys in their reading comprehension of material written with infrequent oral language patterns.

Hypothesis 1 was tested by chi square analysis.For the remaining eight hypotheses, one of two models for the t testwas used depending on whether or not the population variances could be assumed equal.

108 Results In Chapter III, the results of tests of the hypotheses were presented and discussed; measures of reliability and validity of the two instruments devised by the investigator were also presented. The findings were grouped to provide answers to the three questions. With respect to the first question, the chi square analysis of Hypothesis 1 was reported, where 1 degree of freedom was used and the .05 level of significance was accepted. For the total group of 300 subjects the null hypothesis was rejected. For all the subgroups but onethe fourth-grade boysa significantly larger number of subjects obtained their higher reading comprehension score on Test A (frequent oral language patterns) than on Test B (infrequent oral language patterns); subgroupswere defined by grade and by 'sex within and across grades. Derived from Question 2, Hypotheses 2 and 3 dealt with the relation-, ship of grade level and the reading comprehensionscores on Test A and Test B. t tests were used to determine whetheror not obtained differences in mean scores were significant at the .05 level. Since significant differences were found, Hypotheses 2 and 3 were rejected. Hypotheses 4-9 which related to Question 3, focusedon the relationship of sex and reading comprehensionscores on Test A and Test B. To test the differences between mean scores of boys and girlson the two tests, t tests were used. Based on results of the t tests, Hypotheses 5-9 were accepted and Hypothesis 4 was rejected. The results of all the tests of hypotheses were summarized in Table 22 in Chapter III. From the discussion of the reliability and validity of the twomeas- uring instruments, it was concluded that for thegroup of subjects as a whole, and for the second grade, the testswere adequately reliable instruments; reliability of both tests was considerably lower for the fourth

109 grade. The tests seemed to be constructed accordingto valid, logical- constructs based on learning theory and researchon the relationship between oral language ability and reading ability'

Limitations

These limitations should be considered when evaluating thecon- whiCh

The samplecif oral languagepatterns of second and fourth graders that comprised the two reading comprehensiontests devised by the investigator was not eXhaustive. The selection of other frequentand infrequent

patterns=could conceivably yield differentresults. 2. -As the sample used in the studywas not randomly selected from the elementary school population of Madison,results should not be gen- eralized beyond the original sample.

3. Different sets of multiple choicepictures were used in each test. The effect of this differenceon the comprehension scores is not known.

4. The problem of comparing results fromdifferent grade levelswas made apparent by the need to selectone of two ways of nonstructing the tests:(1) completely different tests for eachgrade level, or (2) thesame materials for both grades. As discussedin Chapter II, each alternative involved certain difficulties in interpretingthe results.The decision to use the same set of materials for all subjects and thus controlthe same variables across grades may have resultedin more significant differences between grade levels than actually existed. 5. The reading comprehensiontests (Test A and Test B) werecon- sidered reliablegroup measures for the subjects as a whole and for the second graders; the testswere considerably less reliable for the fourth grade subjects.

110 In both Test A and Test B, oral language patterns with T markers (a coordinating conjunction)were joined to form one sentence (e.g.,In Tesi A, 1 2 4 +T and T 1 2 4 = 1 2 4 + 1 2 4). Strickland described individual patterns,not how they were joined; thus it is not known how closely the resulting combinations resembled children'suse of these patterns.

Conclusions

The following conclusions appear to be warranted withinthe limitations -of the study:

1. A. significant number of second and fourth graderscomprehend material written with frequent oral languagepatterns better than material written with infrequent oral languagepatterns. 2. Fourth graders comprehend materialwritten with frequent and infrequent oral language patterns significantlybetter than second graders.

3. Second- and fourth-grade girls comprehendmaterial written with frequent and infrequent oral languagepatterns better than boys in the same grades,but in general this difference doesnot reach an acceptable level of significance (a= .05). 4. Easily administered, reliable, and validinstruments can be constructed from a relatively small sample of children'soral language patterns to deterznine whetheror not certain patterns are more difficult for children to comprehend than others.

Implications

Implications of the findings for researchin the following areas are discussed: beginning readinginstruction, teacher training,classroom

111 materials, and measurement. More specific research problems that extend from the present study are also outlined.

Implications for Beginning Reading Instruction A developmental pattern was suggested by the finding that the difference between reading comprehension of frequent and infrequentoral patterns seemed-to be greater for seccnd graders than fourth graders. With even younger readers, therefore, the distinction between these divergent language patterns could be expected to be greater. For beginning readers, it seems logical and in keeping with linguistic knowledge to use children's patterns of language structure in written material to facilitate learning the concept that spoken and written language are related. Structuring material more like the way children speak should ease children into this new language skill more efficiently and easily than the use of material which focuses on vocabulary control to the exclusion of the kinds of language structures in which the vocabulary is used. Proponents of language-experience approaches to teaching reading have used children's dictated speech for beginning instruction, but the concept of familiar and unfamiliar language patterns has not been crucial or even evident in these programs. Typically, as soon as the child "gets" the idea of reading, he is moved into trade books or other prepared mate- - rials without further concern for his ability to deal with specific types of sentences. In contrast to this practice, the findings of the present study suggest that children would benefit from some kind of control over sentence pattems until they are readily able to untangle word relationships in any number of infrequent patterns.

Implications for Teacher Training_ The application of linguistics to the teaching of reading has been a relatively new process beginning, perhaps,with Bloomfield and Barnhart' s

112 publication Let's Read (1961).Findings of the present study support linguists who suggest that reading has something to do with the ability to comprehend the large units of language patternsas well as the small units of individual words (Lefevre, 1964).Linguists point out that traditional reading methodology is not rigorously concerned with knowledge aboutour language and its structure (Carroll, 1964). Extension of their principles with reference to the findings of the present study would indicate that teachers would be more effective in helping their children learnto read if they were aware of the following: 1. There is evidence that growth in oral language facility is related to groWth in general reading achievement. The findings of thepresent study do not serve to clarify this position beyond indicating thatcertain oral language patterns are easier to comprehendin written material and that more experienced readers (fourth graders)do not differentiate their reading comprehension of these patternsas much as less experienced readers (second graders). The implication is that teachersshould provide many opportunities for oral language growth, the assumption being that encountering patterns of structure in written materialmay be easier if these patterns are already ina child's repertoire. 2. Teachers should be familiar withsome of the content of linguistics. They should understand the characteristics ofour language--its patternsits regularities, and its irregularitiesand how these affectthe relationship betweenspoken and written language. By, really listening to children, teachers could detecttwo possible sources of difficulty in reading comprehension. First, they could determine to what-extent a -child's dialect deviated from the dialecthe was asked to read; major discrepancies could beexpected to interfere with his understanding of the relationship betweenspoken and written language. Second, teachers could hear whethera child's oral reading reflected his ;- 'understanding:ofWhole groups of wordsintonation patternsrat 'per than single words strung'together. 113 Implications for Classroom Materials

When maximum comprehensionis the goal, writers of materialsfor elementary school children shouldconsider the use of languagepatterns that children find easier tocomprehend than others. Strickland'slist (19 62) of frequent andinfrequent oral languagepatterns, according to the findings of the present study,could provide a general guideto those patterns that should be relativelyeasy or difficult to, comprehend. The guide would be modified, ofcourse, and tempered by the specific natureof the materials,the grade level of the childrenbeing considered, and theintent of the material. The findingsof this study imply thatcontrol over vocabulary is not the onlylogical and desirable controlwhen comprehension of language structuresis essential.

Implications for Measurement ofLanguage- Related Abilities

From the instruments constructedfor this study, it is clearthat tests can be devised to measure the relativedifficulty of select orallanguage patterns in written materials forelementary school children.Several interesting and worthwhilepossibilities are consequentlyavailable for a variety of situations: 1. Since differencesbetween reading comprehensionof frequent and infrequent oral languagepatterns can be measured,a series of tests could be constructed to determinespecific languagepatterns that aremore or less difficult than others.Establishing a hierarchy ofdifficulty could be helpful in thearea of readability as wellas in the area of developmental reading instruction. 2.In the present study itwas felt that determining readingcomprehension of individual,unrelated sentenceswas a practical and relatively

114 uncomplicated procedure. By thissame method, a pool of unrelated test items could be accumulated after pilot testingwith a variety of grade levels and types of children. Theseitems could be grouped in several ways,depending on the intent of the measuringinstrument.For example, by using a sufficiently large number ofitems that dealt with thesame language pattern, it could be determined whetheror not an individual was having difficulty with the pattern. Or,by including a mixture of items labeled "frequent"- and "infrequent,"a survey could be conducted ofan individual's ability to handle divergentlanguage patterns. Construction of such tests could be helpful in deterritiningwhere non-English-speaking children on the one hand, and speakersof non-standard dialectson the other were having-difficulty in"straightening out" the relationships of words in English sentences.

Implications for Research

During the preparation andexecution of this investigation, several research problems emerged thatseem relevant and feasible in light of the findings of this study: 1. An experimental study is neededto determine if the immediateuse of materials structuredmore like the way children speak ismore beneficial to beginning readers than materialswhich do not controlsentence patterns in this manner. The studyshould be longitudinal innature, as long-term effects of suchan innovative program would be importantin terms of atti- tudes as wellas reading skills. 2.Exploratory studies like thisone and Ruddell's (1963) have not dealt with thereason for a gap between children's abilityto speak complex structures and their abilityto read these same structures.Though not a direct outgrowth from thefindings of this study, researchthat would

115 illuminate the cognitiveprocess involved in comprehendinglanguage structures is needed. Withan increased understanding of the interde- pendence of speaking, reading,and writing, teachingprocedures could become better suited to helpingchildren grow successfullyin each of these areas; methods ofassessing progress could becomemore relevant and accurate. 3. More information is neededabout the ability of childrenat various levels of achievement to comprehendselect oral languagepatterns in written material. Discovering thepresence of distinctly different levels of performanceamong good and poor readers at differentgrade levels would help to clarify the generalfindings of the presentstudy. 4. Without the time limitationof the present ztudy,parallel forms of tests similar to TestA and Test B could bedeveloped for different grade levels. These tests wouldprovide a yardstick againstwhich a child's present level of performance couldbe compared to (1) hisachievement the preceding year (ona test constructed for that grade level),or (2) to the achievement of other childrenwithin the same grade.Construction ofa series of parallel tests wouldbe a first step in defining"typical" ability to comprehend divergent languagepatterns at various levels ofdevelop- ment in the elementary school. 5. Without sacrificing thecontrols imposed in TestA and Test B, it should be possible to selectfrequent and infrequent orallanguage patterns so that the same multiple choicepictures could be usedto measure reading comprehension of both kindsof patterns.The use of identicalpictures from test to test would providea tighter control over the instruments.

116 .

APPENDIX A

The. Twenty-five Most Frequent Oral Language Patterns of Second and Fourth Graders as Identified by Strickland (1962)

117 GRADE TWO ORAL LANGUAGE PATTERNS

Rank Order Pattern Frequency

1 1 2 4 309* 2 T 1 2 4 264* 3 1 2 4 + T 189* 4 1 2b 5 130* 5 T 1 2b 5 107

6 T 1 2 4 + T 74* ..., 7 M3 1 2 4 71 8 1 2b 5 + T 57 9 M3 1 2 4 + T 52 10 12 48

11 I 2 Mi + T ...-- 43 12 1 2 + T 42 13 T 1 2 Mi 38 14 1.2 4 T 37 14 T 1 2 4 MI 37 15 T 1 2 36 16 1 2M1 34 17 T 1 21V12 32** 17 1 2 M2 32** 18 1 2 4 M1 31 19 T 1 2 4 M2 29** 20 M3 1 2 M1 + T 28** 21 M3 1 2 + T 27 22' T 1 2b 5 + T 25 23 T M3 112 4 23 21 1 2 3 4 23** 24 T 1 2 3 4 22 24 1 2 4 M2 22 25 1 2 4 Mi + T 21**

* Patterns used in Test A (frequent oral language patterns) ** Patterns used in Test B (infrequent oral language patterns) GRADE FOUR ORAL LANGUAGE PATTERNS

Rank Order Pattern Frequency

1 1 24 202*. 2 T 1 2 4 156*. 3 1 2 4 + T 123*. 4 1 2b 5 108*. 5 T 1 2 4 + T 84*

6. 1 2 4 T 64 7 T 1 2b 5 63,. 8 1 2 Mi + T 51. 9 T 1 2 M1 41 9 T 1 2 4 T 41

10 1 2M 37. . 11 T M3 1 2 4 + T 33 12 1 2b 5 + T 31 13 T 1 2 29 13 M3 1 2 4 29.

14 M3 1 2 4 + T 28 14 1 2 4 M1 28. 15 1 2 4 M2 27. 16 T M3 1 2 4 26. 17 T 1 2 4 Mi + T 25

17 : T 1 2 M2 25** 17 .. M3 1 2 M1 25.- 18 i 12 23 18 T 1 2 4M1 23 18 T 1 2 + T 23

19 1 2 M2 22** 19 T 1 2b 5 + T 22 20 1 2 4 M3 21 20 T 1 2 4 M2 21** 21 M3 1 2 Mi + T 19**

(Continued)

120 GRADE FOUR ORAL LANGUAGE PATTERNS (Continued)

Rank Order Pattern Frequency

21. T 2 4 19 22 1 2 MI. T 18 23 1 2 T 17 23 12M2+T 17 23 1 2 4 Mi + T 17**

24 1 2 3 4 15**. . 24 T M3 1 2 Mi 15 25 T 1 2 4 14

* Patterns usedin Test A (frequent oral language patterns) ** Patterns usedin Test B (infrequent oral language patterns) .Patterns usedin Ruddell's high frequency passages (1963) ..Patterns usedin Ruddell's low frequency passages (1963)

121 APPENDIX B

The Measuring Instruments: Directions and Reading Comprehension Tests A and B Appendix B is not printed herein. See original manuscript in the Uni- versity of Wisconsin Memorial Library..

Pal 123 APPENDIX C

Language Patterns in the Sentences of TestA and Test B Appendix C is not printed herein. Seeoriginal manuscript in the Uni- versity of Wisconsin Memorial Library. APPENDIX D

Reading Comprehension Test Scores for All Subjects Appendix D is not printed herein. See original manuscript in the Uni- versity of Wisconsin Memorial Library.

124127 APPENDIX E

Frequency Distributions of Scoreson Test A and Test B Within and Across Grades Two and Four 2224 I,c,..) 181020 e si---too-- Fourth-Grade - --o Second-Grade Subjects Subjects SI 1 1 1214 / 1 /0- -0 / / 10 1 846 / 2 8 Frequency Distributions10 of Total Scores on 12 14 16 TOTAL18 TEST SCORE: TEST A 20 22 Test A for Second and24 Fourth Grades 26 28 30 32 34 36 38 24 - 16182022 - - 0-11------0 Fourth-Grade Second-Grade Subjects Subjects 101412 - 8 - .- 4 26- - sclP 0' / / 10 12 -4)--0, 14 .0--16 18 20 22 A / 24V t, 26 28 I 30 32 34J 36 f Frequency Distributions of Total Scores on Test B for Second and Fourth Grades TOTAL TEST SCORE: TEST B 384042 . 323436 -. 3028 -. (>...... 0...... All Subjects: Test A All Subjects: Test B 222426 -- 161820 - 101214 . 426,8. 10Frequency Distribution of Total 12 14 16 18 TOTAL TEST SCORES 20 Scores for All Subjects 22 24 26 on Test28 A and Test B 30 32 34 36 38 APPENDIX F

Item Analysis of Test A andTest B for Grades TWo andFour Appendix F is not printedherein. See originalmanuscript in the Uni- versity of Wisconsin MemorialLibrary.

)

116 137 BIBLIOGRAPHY

Amsden, Constance E."Oral Language and Printed-Word Reading." Claremont College Reading Conference Twenty-ninthYearbook. Claremont, California: Claremont College CurriculumLaboratory, 1964, pp. 90-3. Anastasi, Anne. Psychological Testing. New York: TheMacmillan Company, 1961. Baker, Frank B. and T. J. Martin." FORTAP, A Fortran TestAnalysis Package. The University of Wisconsin: Wisconsin Researchand Development Center for Cognitive Learning, March1, 1968. Barrett, Thomas C. The Relationship Between Selected Reading Readiness Measures of Visual Discrimination and First Grade ReadingAchieve- ment. Unpublished doctoral dissertation.University of Minnesota, 1962. Barrett, Thomas C. and Wayne Otto."Elementary Pupil's Ability to Con- ceptualize the Main Idea in Reading." Apaper presented at the American Educational Research Association meeting, Chicago,Feb- ruary, 1968. Betts,Emmett Albert."Readability: Its Application to the Elementary School." Journal of Educational Research. 42:6, February,1949, pp. 439-59. Bloom, Benjamin. "Test Reliability for What?" Readingsin Educational and Psychological Measurement.(ed. by Clinton Chase and H. Glenn Ludlow), New York: Houghton Mifflin Company,1966, pp. 126-34. Bloomfield, Leonard and Clarence. Barnhart. Let'sRead. Detroit: Wayne State University Press, 1961. Burrows,Alvina. "Research Critiques." Elementary English.41, May, 1964, pp. 532-5. Carroll, John B. "The Analysis of Reading Instruction: Perspectives from Psychology and Linguistics ," Theories of Learning and Instruction (ed. by Ernest R. Hilgard), Chicago, Illinois: The National Society for the Study of Education, 1964, pp. 336-53. ."Communication Theory, Linguistics, and Psycholinguistics," Review of Educational Research.28:2, April, 1958. "Language Development," Encyclopedia of Educational Research. (ed. by Chester W. Harris), New York: The Macmillan Company, 1960, pp. 283-92. "Linguistics and the Psychology of Language," Review of Educa-

tional Research. 34:2; April, 1964. . Cronbach,. Lee 1.Essentials of Psychological Testing. New York: Harper and Brothers, Publishers, 1960. Davis, 0.. L., Jr. and Joan G. Seifert. "Some Linguistic Features of Five Literature Book for Children." Elementary English. December, 1967. Ebel, Robert. In Testing Problems in Perspective.(ed. by Anne Anastasi) American Council on Education, 1966, pp. 115-30. Fries,Charles. Linailistics and Reading. New York: Holt, Rinehart and .Brothers, Publishers, 1963. Goodman, Kenneth. Stud of Children's Behavior While Readin Orall U. S. Department of Health, Education, and Welfare, Office of Edu- cation, Project No. S 425, Detroit, Michigan: Wayne State University, 1968. Gray, William S. "Progress in the Study of Readability." The Elementary -School journal. May, 1947, pp. 491-99. Guha, Susanta. "Multiple Regression Analysis REGAL," The University of Wisconsin Computing Center, Users Manual, Vol. 6 - Preliminary, Statistical Programs for the 1604 Computer. Nov. 1966, Section 3.2. Hays,William L. Statistics for Psychologists. New York: Holt, Rine_- hart and Winston, 1963. Hooker, Mary Elsa. Reading Materials for Children Based on Their Lan- guage Patterns oiSyntax, Vocabulary( and Interests. Unpublished Master' s thesis,University of Arizona, 1963. Hodge, Robert, Paul Siegel, and Peter Rossi. "Occupational Prestige in in the United States, 1925 to 1963," American Journal of Sociology, 70:286-302, November,1964.

140 Hunt, Kellogg W. "Recent Measures in Language Development." A paper presented at the National Conference on Research in English, Chicago, February, 1966. Grammatical Structures Written at Three Grade Levels. U. S. Department of Health, Education, and Welfare, Office of Education, Cooperative Research Project No. 1998. Tallahasee, Florida: Florida State University, 1965. Kean, John. The Linguistic Structure of Second and Fifth Grade Teachers' Oral Classroom Language. V. S. Department of Health, Education, and Welfare, Office of Education, Cooperative Research Project No. S-331, Madison, Wisconsin: University of Wisconsin, 1967. Kelley, Truman L.and Richard Madden, Eric Gardner, and Herbert Rudman. Stanford Achievement Test: Intermediate I and Primary II, Form Y. New York: Harcourt, Brace and World, Inc., 1964. Kjeldergaard, Paul M. "The Psychology of Language," Review of Educa- tional Research. 31:2, April, 1961. Lefevre, Carl. Linguistics and the Teaching of Reading. New York: McGraw -Hill Book Company, 1964. Lindgren, Henry Clay. Educational Psychology inthe Classroom. New York: John Wiley and -Sons,Inc.,1962. Lindvall, C. M. Measuring Pupil Achievement and Aptitude. New York: Harcourt, Brace and World, Inc.,1967. .Loban, Walter. The Language of Elementary School Children. National Council of Teachers.of English, Research Report No. 1. Champaign, Illinois: the Council, 1963. Mayer, Edgar. "Research Critiques." Elementary English. 41, May, 1964," pp. 535-36, + 541. McCarthy, Dorothea. "Language Development in Children." A Manual of Child Psychology.(ed. by Leonard Charmichael) New York: John Wiley and Sons; Inc.,1954. Mortenson, William Paul. Differences in Performance of Beginning First. Grade Pupils on Selected Pre-Reading Tasks According to Socio- Economic Status and Sex. Unpublished doctoral dissertation, Uni- versity of Wisconsin, Madison, 1966. 0.'Donnell,i Roy and William Griffin, and Raymond Norris. Syntax of Kin- tenarten and Elements r School Children: A 'Transformational Analysis. National Council of Teachers of English, Research Report No. 8. Champaign, Illinois: the .Council, 1967.

141 Reiss, Albert J. Occupations andSocial Status. New York: The Free Press of Glencoe, Inc. , 1961. Riling, Mildred. Oral and Written Lan ua eof Children in Grades 4 and 6 Porn ared with the Lan ua e of TheirTextbooks. U.S. Department of Health, Education, and Welfare, Office ofEducation, Cooperative Research Project No. 2410. Durant, Oklahoma:Southeastern State College, 1965. Ruddell, Robert Byron. An Investigation ofthe Effect of the Similarity of Oral and Written Patterns of Lan ua e Structure onReadin Com re- hension. Unpublished doctoral dissertation,Indiana University: 1963. .The Effect of Four Programs of ReadingInstruction with Varying Emphasis on the Regularity of Grapheme-PhonemeCorrespondences and the Relation of Lan ua e Structure to Meanin onAchievement in First Grade Reading. U.S. Departmentof Health, Education, and Welfare, Office of Education, Cooperative ResearchProject No. 2699. Berkeley, California: University ofCalifornia, 1965. Stevens, Martin. "Intonation in the Teachingof Reading." Elementary English. 42:231-37, March, 1965. Stone, C. R. "Measuring Difficulty of PrimaryReading Material: A Con- structive Criticism of Spache's Measure."Elementary School Journal. 57:410-13, March, 1953. Strickland, Ruth G. "The Language of ElementarySchool Children: Its Relationship to the Language of Reading Textbooksand the Quality of Reading of Selected Children," Indiana UniversitySchool of Educa- tion Bulletin.38: 1-131, July, 1962. Tate, Merle. Statistics in Education. New York: TheMacmillan Com- pany, 1955. Tatham, Susan M. "The Effect of MaterialWritten with Certain Oral Lan- guage Patterns on ReadingComprehension." A paper presented at the American Educational Research Association meeting,New York City, February, 1967. ."Using Multiple Choice Questions to Measurethe Effect on Com- prehension of Material Written with Select Oral LanguagePatterns." A paper presented at the American EducationalResearch Association meeting, Chicago, February, 1968. Templin, Mildred. "Certain Language Skills in Children:Their Develop- ment and Interrelationships." Institute ofChild Welfare Monograph Series, No. 26. Minneapolis: University, of MinnesotaPress, 1957.

142 Vygotsky, Lev Semenovich. Thought and Language. Cambridge, Mass.: Massachusetts Institute of Technology Press, 1962. Winer, B. J.Statistical Princi les in Ex erimental Desin. New York: McGraw-Hill Book Company, 1962.

143